H7.9 top features

Top feature 0 in H7.9: (feature 34521)

TOP ACTIVATIONS
MAX = 8.629

of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
Jason
Token Jason
Feature activation+0.000
was
Token was
Feature activation+7.033
a
Token a
Feature activation+8.207
member
Token member
Feature activation+8.629
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.000
,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
Jason
Token Jason
Feature activation+0.000
was
Token was
Feature activation+7.033
a
Token a
Feature activation+8.207
member
Token member
Feature activation+8.629
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
Christmas
Token Christmas
Feature activation+0.000
Eve
Token Eve
Feature activation+0.000
,
Token,
Feature activation+0.000
Indians
Token Indians
Feature activation+0.000
,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
Jason
Token Jason
Feature activation+0.000
was
Token was
Feature activation+7.033
a
Token a
Feature activation+8.207
member
Token member
Feature activation+8.629
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
nut
Token nut
Feature activation+0.000
items
Token items
Feature activation+0.551
are
Token are
Feature activation+3.430
a
Token a
Feature activation+5.444
viable
Token viable
Feature activation+5.071
source
Token source
Feature activation+6.528
.
Token.
Feature activation+0.000
There
Token There
Feature activation+0.000
are
Token are
Feature activation+0.000
many
Token many
Feature activation+0.000
nut
Token nut
Feature activation+0.000
Sarah
Token Sarah
Feature activation+0.373
and
Token and
Feature activation+0.000
her
Token her
Feature activation+0.093
unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
I
Token I
Feature activation+5.213
am
Token am
Feature activation+6.344
at
Token at
Feature activation+2.998
the
Token the
Feature activation+3.760
tail
Token tail
Feature activation+1.956
end
Token end
Feature activation+6.364
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
reason
Token reason
Feature activation+0.000
years
Token years
Feature activation+0.000
ago
Token ago
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
I
Token I
Feature activation+5.213
am
Token am
Feature activation+6.344
at
Token at
Feature activation+2.998
the
Token the
Feature activation+3.760
tail
Token tail
Feature activation+1.956
end
Token end
Feature activation+6.364
.
Token.
Feature activation+0.000
her
Token her
Feature activation+0.093
unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
Christmas
Token Christmas
Feature activation+0.000
Eve
Token Eve
Feature activation+0.000
and
Token and
Feature activation+0.000
her
Token her
Feature activation+0.093
unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
Christmas
Token Christmas
Feature activation+0.000
which
Token which
Feature activation+0.000
Dr
Token Dr
Feature activation+0.000
Gol
Token Gol
Feature activation+0.000
ut
Tokenut
Feature activation+0.000
vin
Tokenvin
Feature activation+0.000
is
Token is
Feature activation+6.019
chief
Token chief
Feature activation+5.574
scientist
Token scientist
Feature activation+4.868
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
ot
Tokenot
Feature activation+0.000
ides
Tokenides
Feature activation+3.310
G
Token G
Feature activation+0.000
GG
TokenGG
Feature activation+0.000
are
Token are
Feature activation+5.696
the
Token the
Feature activation+5.726
P
Token P
Feature activation+0.000
AM
TokenAM
Feature activation+0.139
sequence
Token sequence
Feature activation+1.298
(
Token (
Feature activation+0.000
Fig
TokenFig
Feature activation+0.000
nucle
Token nucle
Feature activation+0.000
ot
Tokenot
Feature activation+0.000
ides
Tokenides
Feature activation+3.310
G
Token G
Feature activation+0.000
GG
TokenGG
Feature activation+0.000
are
Token are
Feature activation+5.696
the
Token the
Feature activation+5.726
P
Token P
Feature activation+0.000
AM
TokenAM
Feature activation+0.139
sequence
Token sequence
Feature activation+1.298
(
Token (
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+5.100
are
Token are
Feature activation+5.492
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000
very
Token very
Feature activation+4.840
proud
Token proud
Feature activation+5.329
.
Token.
Feature activation+0.000
country
Token country
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+5.100
are
Token are
Feature activation+5.492
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000
very
Token very
Feature activation+4.840
proud
Token proud
Feature activation+5.329
Dr
Token Dr
Feature activation+0.000
Gol
Token Gol
Feature activation+0.000
ut
Tokenut
Feature activation+0.000
vin
Tokenvin
Feature activation+0.000
is
Token is
Feature activation+6.019
chief
Token chief
Feature activation+5.574
scientist
Token scientist
Feature activation+4.868
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Scientists
TokenScientists
Feature activation+0.000
great
Token great
Feature activation+0.000
country
Token country
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+5.100
are
Token are
Feature activation+5.492
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000
very
Token very
Feature activation+4.840
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
nut
Token nut
Feature activation+0.000
items
Token items
Feature activation+0.551
are
Token are
Feature activation+3.430
a
Token a
Feature activation+5.444
viable
Token viable
Feature activation+5.071
source
Token source
Feature activation+6.528
.
Token.
Feature activation+0.000
There
Token There
Feature activation+0.000
are
Token are
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+5.100
are
Token are
Feature activation+5.492
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000
very
Token very
Feature activation+4.840
proud
Token proud
Feature activation+5.329
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000
very
Token very
Feature activation+4.840
proud
Token proud
Feature activation+5.329
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
you
Token you
Feature activation+0.000
so
Token so
Feature activation+0.000
much
Token much
Feature activation+0.000

Top DFA by src position
MAX = 9.551

ew
Tokenew
Feature activation-0.004
Top resid features:
a
Tokena
Feature activation-0.012
Top resid features:
Indians
Token Indians
Feature activation-0.012
Top resid features:
,
Token,
Feature activation-0.002
Top resid features:
of
Token of
Feature activation+0.078
Top resid features:
which
Token which
Feature activation+9.551
Top resid features:
Jason
Token Jason
Feature activation+0.002
Top resid features:
was
Token was
Feature activation+0.540
Top resid features:
a
Token a
Feature activation+0.316
Top resid features:
member
Token member
Feature activation+0.045
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
ew
Tokenew
Feature activation-0.007
Top resid features:
a
Tokena
Feature activation-0.012
Top resid features:
Indians
Token Indians
Feature activation+0.029
Top resid features:
,
Token,
Feature activation+0.011
Top resid features:
of
Token of
Feature activation+0.108
Top resid features:
which
Token which
Feature activation+9.117
Top resid features:
Jason
Token Jason
Feature activation+0.053
Top resid features:
was
Token was
Feature activation+0.636
Top resid features:
a
Token a
Feature activation+0.087
Top resid features:
member
Token member
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
a
Token a
Feature activation-0.005
Top resid features:
car
Token car
Feature activation-0.013
Top resid features:
crash
Token crash
Feature activation-0.027
Top resid features:
,
Token,
Feature activation-0.087
Top resid features:
of
Token of
Feature activation+0.055
Top resid features:
which
Token which
Feature activation+8.480
Top resid features:
Sarah
Token Sarah
Feature activation+0.003
Top resid features:
and
Token and
Feature activation-0.002
Top resid features:
her
Token her
Feature activation+0.035
Top resid features:
unborn
Token unborn
Feature activation+0.127
Top resid features:
baby
Token baby
Feature activation+0.108
Top resid features:
ew
Tokenew
Feature activation-0.002
Top resid features:
a
Tokena
Feature activation-0.007
Top resid features:
Indians
Token Indians
Feature activation+0.008
Top resid features:
,
Token,
Feature activation-0.007
Top resid features:
of
Token of
Feature activation+0.137
Top resid features:
which
Token which
Feature activation+8.541
Top resid features:
Jason
Token Jason
Feature activation+0.083
Top resid features:
was
Token was
Feature activation+0.172
Top resid features:
a
Token a
Feature activation+0.000
Top resid features:
member
Token member
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
foods
Token foods
Feature activation-0.038
Top resid features:
that
Token that
Feature activation-0.102
Top resid features:
contain
Token contain
Feature activation-0.023
Top resid features:
protein
Token protein
Feature activation-0.055
Top resid features:
of
Token of
Feature activation+0.076
Top resid features:
which
Token which
Feature activation+7.602
Top resid features:
nut
Token nut
Feature activation-0.064
Top resid features:
items
Token items
Feature activation+0.085
Top resid features:
are
Token are
Feature activation+0.715
Top resid features:
a
Token a
Feature activation+0.215
Top resid features:
viable
Token viable
Feature activation+0.097
Top resid features:
a
Token a
Feature activation-0.008
Top resid features:
car
Token car
Feature activation-0.016
Top resid features:
crash
Token crash
Feature activation-0.099
Top resid features:
,
Token,
Feature activation-0.053
Top resid features:
of
Token of
Feature activation+0.047
Top resid features:
which
Token which
Feature activation+7.718
Top resid features:
Sarah
Token Sarah
Feature activation+0.015
Top resid features:
and
Token and
Feature activation+0.040
Top resid features:
her
Token her
Feature activation+0.099
Top resid features:
unborn
Token unborn
Feature activation+0.202
Top resid features:
baby
Token baby
Feature activation+0.341
Top resid features:
began
Token began
Feature activation-0.029
Top resid features:
six
Token six
Feature activation+0.002
Top resid features:
years
Token years
Feature activation-0.009
Top resid features:
ago
Token ago
Feature activation-0.060
Top resid features:
of
Token of
Feature activation+0.012
Top resid features:
which
Token which
Feature activation+7.336
Top resid features:
I
Token I
Feature activation+0.109
Top resid features:
am
Token am
Feature activation+0.529
Top resid features:
at
Token at
Feature activation+0.108
Top resid features:
the
Token the
Feature activation+0.036
Top resid features:
tail
Token tail
Feature activation-0.180
Top resid features:
began
Token began
Feature activation-0.019
Top resid features:
six
Token six
Feature activation+0.000
Top resid features:
years
Token years
Feature activation-0.014
Top resid features:
ago
Token ago
Feature activation-0.059
Top resid features:
of
Token of
Feature activation+0.010
Top resid features:
which
Token which
Feature activation+7.870
Top resid features:
I
Token I
Feature activation+0.191
Top resid features:
am
Token am
Feature activation+0.071
Top resid features:
at
Token at
Feature activation+0.000
Top resid features:
the
Token the
Feature activation+0.000
Top resid features:
tail
Token tail
Feature activation+0.000
Top resid features:
a
Token a
Feature activation-0.003
Top resid features:
car
Token car
Feature activation-0.002
Top resid features:
crash
Token crash
Feature activation-0.026
Top resid features:
,
Token,
Feature activation-0.024
Top resid features:
of
Token of
Feature activation+0.023
Top resid features:
which
Token which
Feature activation+6.791
Top resid features:
Sarah
Token Sarah
Feature activation+0.091
Top resid features:
and
Token and
Feature activation-0.007
Top resid features:
her
Token her
Feature activation+0.023
Top resid features:
unborn
Token unborn
Feature activation+0.133
Top resid features:
baby
Token baby
Feature activation+0.267
Top resid features:
a
Token a
Feature activation-0.001
Top resid features:
car
Token car
Feature activation-0.012
Top resid features:
crash
Token crash
Feature activation-0.044
Top resid features:
,
Token,
Feature activation-0.014
Top resid features:
of
Token of
Feature activation+0.045
Top resid features:
which
Token which
Feature activation+7.243
Top resid features:
Sarah
Token Sarah
Feature activation+0.017
Top resid features:
and
Token and
Feature activation+0.016
Top resid features:
her
Token her
Feature activation+0.159
Top resid features:
unborn
Token unborn
Feature activation+0.048
Top resid features:
baby
Token baby
Feature activation+0.171
Top resid features:
L
Token L
Feature activation+0.004
Top resid features:
HC
TokenHC
Feature activation-0.006
Top resid features:
b
Tokenb
Feature activation+0.009
Top resid features:
,
Token,
Feature activation+0.008
Top resid features:
of
Token of
Feature activation+0.139
Top resid features:
which
Token which
Feature activation+7.387
Top resid features:
Dr
Token Dr
Feature activation-0.019
Top resid features:
Gol
Token Gol
Feature activation+0.014
Top resid features:
ut
Tokenut
Feature activation+0.010
Top resid features:
vin
Tokenvin
Feature activation-0.005
Top resid features:
is
Token is
Feature activation+0.046
Top resid features:
asm
Tokenasm
Feature activation-0.001
Top resid features:
id
Tokenid
Feature activation+0.007
Top resid features:
target
Token target
Feature activation-0.019
Top resid features:
sequence
Token sequence
Feature activation+0.014
Top resid features:
of
Token of
Feature activation+0.053
Top resid features:
which
Token which
Feature activation+6.071
Top resid features:
the
Token the
Feature activation+0.110
Top resid features:
nucle
Token nucle
Feature activation-0.012
Top resid features:
ot
Tokenot
Feature activation-0.011
Top resid features:
ides
Tokenides
Feature activation+0.121
Top resid features:
G
Token G
Feature activation+0.050
Top resid features:
asm
Tokenasm
Feature activation-0.004
Top resid features:
id
Tokenid
Feature activation+0.010
Top resid features:
target
Token target
Feature activation-0.021
Top resid features:
sequence
Token sequence
Feature activation+0.049
Top resid features:
of
Token of
Feature activation+0.041
Top resid features:
which
Token which
Feature activation+6.600
Top resid features:
the
Token the
Feature activation+0.166
Top resid features:
nucle
Token nucle
Feature activation+0.004
Top resid features:
ot
Tokenot
Feature activation-0.010
Top resid features:
ides
Tokenides
Feature activation+0.160
Top resid features:
G
Token G
Feature activation+0.042
Top resid features:
to
Token to
Feature activation-0.007
Top resid features:
this
Token this
Feature activation+0.003
Top resid features:
great
Token great
Feature activation-0.022
Top resid features:
country
Token country
Feature activation-0.002
Top resid features:
of
Token of
Feature activation-0.013
Top resid features:
which
Token which
Feature activation+6.625
Top resid features:
we
Token we
Feature activation+0.131
Top resid features:
are
Token are
Feature activation+0.328
Top resid features:
all
Token all
Feature activation+0.273
Top resid features:
so
Token so
Feature activation+0.033
Top resid features:
very
Token very
Feature activation+0.000
Top resid features:
to
Token to
Feature activation-0.010
Top resid features:
this
Token this
Feature activation+0.010
Top resid features:
great
Token great
Feature activation-0.064
Top resid features:
country
Token country
Feature activation-0.006
Top resid features:
of
Token of
Feature activation-0.023
Top resid features:
which
Token which
Feature activation+6.865
Top resid features:
we
Token we
Feature activation+0.057
Top resid features:
are
Token are
Feature activation+0.356
Top resid features:
all
Token all
Feature activation+0.122
Top resid features:
so
Token so
Feature activation+0.000
Top resid features:
very
Token very
Feature activation+0.000
Top resid features:
L
Token L
Feature activation+0.002
Top resid features:
HC
TokenHC
Feature activation-0.010
Top resid features:
b
Tokenb
Feature activation+0.011
Top resid features:
,
Token,
Feature activation+0.007
Top resid features:
of
Token of
Feature activation+0.063
Top resid features:
which
Token which
Feature activation+6.581
Top resid features:
Dr
Token Dr
Feature activation+0.018
Top resid features:
Gol
Token Gol
Feature activation-0.021
Top resid features:
ut
Tokenut
Feature activation+0.003
Top resid features:
vin
Tokenvin
Feature activation-0.014
Top resid features:
is
Token is
Feature activation+0.414
Top resid features:
to
Token to
Feature activation-0.032
Top resid features:
this
Token this
Feature activation+0.011
Top resid features:
great
Token great
Feature activation-0.066
Top resid features:
country
Token country
Feature activation-0.027
Top resid features:
of
Token of
Feature activation-0.058
Top resid features:
which
Token which
Feature activation+7.120
Top resid features:
we
Token we
Feature activation+0.163
Top resid features:
are
Token are
Feature activation+0.109
Top resid features:
all
Token all
Feature activation+0.000
Top resid features:
so
Token so
Feature activation+0.000
Top resid features:
very
Token very
Feature activation+0.000
Top resid features:
foods
Token foods
Feature activation-0.023
Top resid features:
that
Token that
Feature activation-0.016
Top resid features:
contain
Token contain
Feature activation-0.001
Top resid features:
protein
Token protein
Feature activation-0.065
Top resid features:
of
Token of
Feature activation+0.085
Top resid features:
which
Token which
Feature activation+6.600
Top resid features:
nut
Token nut
Feature activation-0.093
Top resid features:
items
Token items
Feature activation+0.119
Top resid features:
are
Token are
Feature activation+0.934
Top resid features:
a
Token a
Feature activation+0.079
Top resid features:
viable
Token viable
Feature activation+0.000
Top resid features:
to
Token to
Feature activation-0.014
Top resid features:
this
Token this
Feature activation+0.004
Top resid features:
great
Token great
Feature activation-0.036
Top resid features:
country
Token country
Feature activation-0.001
Top resid features:
of
Token of
Feature activation-0.013
Top resid features:
which
Token which
Feature activation+6.632
Top resid features:
we
Token we
Feature activation+0.111
Top resid features:
are
Token are
Feature activation+0.211
Top resid features:
all
Token all
Feature activation+0.224
Top resid features:
so
Token so
Feature activation-0.027
Top resid features:
very
Token very
Feature activation+0.024
Top resid features:
to
Token to
Feature activation-0.014
Top resid features:
this
Token this
Feature activation+0.004
Top resid features:
great
Token great
Feature activation-0.042
Top resid features:
country
Token country
Feature activation+0.010
Top resid features:
of
Token of
Feature activation-0.067
Top resid features:
which
Token which
Feature activation+6.472
Top resid features:
we
Token we
Feature activation+0.143
Top resid features:
are
Token are
Feature activation+0.158
Top resid features:
all
Token all
Feature activation+0.147
Top resid features:
so
Token so
Feature activation+0.040
Top resid features:
very
Token very
Feature activation+0.072
Top resid features:

Decoder Weights Distribution

Head 0: 0.05

Head 1: 0.03

Head 2: 0.01

Head 3: 0.06

Head 4: 0.05

Head 5: 0.04

Head 6: 0.03

Head 7: 0.01

Head 8: 0.14

Head 9: 0.55

Head 10: 0.01

Head 11: 0.02

Positive logits

alleg2.05

rals2.02

��1.97

ruct1.96

1.93

--------1.93

��1.93

arer1.91

Attribute1.91

Hood1.84

ccess1.83

quished1.80

Mod1.80

ember1.79

)1.78

Them1.77

ECD1.76

inction1.75

Marsh1.73

orig1.72

Negative logits

freezes-2.49

freezing-2.17

nep-2.10

unda-1.92

FANTASY-1.92

eli-1.90

ilege-1.89

chi-1.89

acha-1.89

freeze-1.84

Solitaire-1.81

nai-1.78

stabilized-1.76

ga-1.73

uana-1.73

smoot-1.72

rolet-1.72

froze-1.69

adobe-1.68

paren-1.68

INTERVAL 7.766 - 8.629
CONTAINS 0.000%

,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
Jason
Token Jason
Feature activation+0.000
was
Token was
Feature activation+7.033
a
Token a
Feature activation+8.207
member
Token member
Feature activation+8.629
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
Jason
Token Jason
Feature activation+0.000
was
Token was
Feature activation+7.033
a
Token a
Feature activation+8.207
member
Token member
Feature activation+8.629
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.000

INTERVAL 6.903 - 7.766
CONTAINS 0.000%

unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
Christmas
Token Christmas
Feature activation+0.000
Eve
Token Eve
Feature activation+0.000
,
Token,
Feature activation+0.000
Indians
Token Indians
Feature activation+0.000
,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
Jason
Token Jason
Feature activation+0.000
was
Token was
Feature activation+7.033
a
Token a
Feature activation+8.207
member
Token member
Feature activation+8.629
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000

INTERVAL 6.040 - 6.903
CONTAINS 0.001%

nut
Token nut
Feature activation+0.000
items
Token items
Feature activation+0.551
are
Token are
Feature activation+3.430
a
Token a
Feature activation+5.444
viable
Token viable
Feature activation+5.071
source
Token source
Feature activation+6.528
.
Token.
Feature activation+0.000
There
Token There
Feature activation+0.000
are
Token are
Feature activation+0.000
many
Token many
Feature activation+0.000
nut
Token nut
Feature activation+0.000
Sarah
Token Sarah
Feature activation+0.373
and
Token and
Feature activation+0.000
her
Token her
Feature activation+0.093
unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
I
Token I
Feature activation+5.213
am
Token am
Feature activation+6.344
at
Token at
Feature activation+2.998
the
Token the
Feature activation+3.760
tail
Token tail
Feature activation+1.956
end
Token end
Feature activation+6.364
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
reason
Token reason
Feature activation+0.000
her
Token her
Feature activation+0.093
unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
Christmas
Token Christmas
Feature activation+0.000
Eve
Token Eve
Feature activation+0.000
and
Token and
Feature activation+0.000
her
Token her
Feature activation+0.093
unborn
Token unborn
Feature activation+0.000
baby
Token baby
Feature activation+0.000
were
Token were
Feature activation+6.492
the
Token the
Feature activation+6.195
only
Token only
Feature activation+6.309
survivors
Token survivors
Feature activation+7.712
.
Token.
Feature activation+0.000
On
Token On
Feature activation+0.000
Christmas
Token Christmas
Feature activation+0.000

INTERVAL 5.177 - 6.040
CONTAINS 0.001%

nucle
Token nucle
Feature activation+0.000
ot
Tokenot
Feature activation+0.000
ides
Tokenides
Feature activation+3.310
G
Token G
Feature activation+0.000
GG
TokenGG
Feature activation+0.000
are
Token are
Feature activation+5.696
the
Token the
Feature activation+5.726
P
Token P
Feature activation+0.000
AM
TokenAM
Feature activation+0.139
sequence
Token sequence
Feature activation+1.298
(
Token (
Feature activation+0.000
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000
very
Token very
Feature activation+4.840
proud
Token proud
Feature activation+5.329
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
you
Token you
Feature activation+0.000
so
Token so
Feature activation+0.000
much
Token much
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+5.100
are
Token are
Feature activation+5.492
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000
very
Token very
Feature activation+4.840
proud
Token proud
Feature activation+5.329
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
nut
Token nut
Feature activation+0.000
items
Token items
Feature activation+0.551
are
Token are
Feature activation+3.430
a
Token a
Feature activation+5.444
viable
Token viable
Feature activation+5.071
source
Token source
Feature activation+6.528
.
Token.
Feature activation+0.000
There
Token There
Feature activation+0.000
are
Token are
Feature activation+0.000
six
Token six
Feature activation+0.000
years
Token years
Feature activation+0.000
ago
Token ago
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
I
Token I
Feature activation+5.213
am
Token am
Feature activation+6.344
at
Token at
Feature activation+2.998
the
Token the
Feature activation+3.760
tail
Token tail
Feature activation+1.956
end
Token end
Feature activation+6.364

INTERVAL 4.314 - 5.177
CONTAINS 0.002%

which
Token which
Feature activation+0.000
nut
Token nut
Feature activation+0.000
items
Token items
Feature activation+0.551
are
Token are
Feature activation+3.430
a
Token a
Feature activation+5.444
viable
Token viable
Feature activation+5.071
source
Token source
Feature activation+6.528
.
Token.
Feature activation+0.000
There
Token There
Feature activation+0.000
are
Token are
Feature activation+0.000
many
Token many
Feature activation+0.000
Gol
Token Gol
Feature activation+0.000
ut
Tokenut
Feature activation+0.000
vin
Tokenvin
Feature activation+0.000
is
Token is
Feature activation+6.019
chief
Token chief
Feature activation+5.574
scientist
Token scientist
Feature activation+4.868
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Scientists
TokenScientists
Feature activation+0.000
hope
Token hope
Feature activation+0.000
colonial
Token colonial
Feature activation+0.000
officials
Token officials
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+1.282
people
Token people
Feature activation+4.523
of
Token of
Feature activation+0.000
that
Token that
Feature activation+0.000
area
Token area
Feature activation+0.524
had
Token had
Feature activation+4.456
had
Token had
Feature activation+4.244
midfield
Token midfield
Feature activation+0.000
,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
there
Token there
Feature activation+2.245
are
Token are
Feature activation+4.453
many
Token many
Feature activation+4.887
like
Token like
Feature activation+0.000
it
Token it
Feature activation+0.000
."
Token."
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
this
Token this
Feature activation+0.000
great
Token great
Feature activation+0.000
country
Token country
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+5.100
are
Token are
Feature activation+5.492
all
Token all
Feature activation+5.607
so
Token so
Feature activation+5.652
very
Token very
Feature activation+5.341
,
Token,
Feature activation+0.000

INTERVAL 3.452 - 4.314
CONTAINS 0.001%

of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
I
Token I
Feature activation+5.213
am
Token am
Feature activation+6.344
at
Token at
Feature activation+2.998
the
Token the
Feature activation+3.760
tail
Token tail
Feature activation+1.956
end
Token end
Feature activation+6.364
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+2.888
here
Token here
Feature activation+1.966
on
Token on
Feature activation+0.000
Earth
Token Earth
Feature activation+2.611
have
Token have
Feature activation+3.981
never
Token never
Feature activation+3.473
even
Token even
Feature activation+2.889
heard
Token heard
Feature activation+4.350
.
Token.
Feature activation+0.000
Close
Token Close
Feature activation+0.000
we
Token we
Feature activation+2.888
here
Token here
Feature activation+1.966
on
Token on
Feature activation+0.000
Earth
Token Earth
Feature activation+2.611
have
Token have
Feature activation+3.981
never
Token never
Feature activation+3.473
even
Token even
Feature activation+2.889
heard
Token heard
Feature activation+4.350
.
Token.
Feature activation+0.000
Close
Token Close
Feature activation+0.000
to
Token to
Feature activation+0.000
people
Token people
Feature activation+4.523
of
Token of
Feature activation+0.000
that
Token that
Feature activation+0.000
area
Token area
Feature activation+0.524
had
Token had
Feature activation+4.456
had
Token had
Feature activation+4.244
the
Token the
Feature activation+2.424
most
Token most
Feature activation+3.066
experience
Token experience
Feature activation+4.613
,"
Token,"
Feature activation+0.000
he
Token he
Feature activation+0.000
which
Token which
Feature activation+0.000
his
Token his
Feature activation+2.449
or
Token or
Feature activation+0.000
her
Token her
Feature activation+3.089
purchase
Token purchase
Feature activation+2.398
is
Token is
Feature activation+3.663
a
Token a
Feature activation+2.966
part
Token part
Feature activation+4.623
,
Token,
Feature activation+0.000
the
Token the
Feature activation+0.000
probability
Token probability
Feature activation+0.000

INTERVAL 2.589 - 3.452
CONTAINS 0.002%

of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.450
nucle
Token nucle
Feature activation+0.000
ot
Tokenot
Feature activation+0.000
ides
Tokenides
Feature activation+3.310
G
Token G
Feature activation+0.000
GG
TokenGG
Feature activation+0.000
are
Token are
Feature activation+5.696
the
Token the
Feature activation+5.726
P
Token P
Feature activation+0.000
specified
Token specified
Feature activation+0.000
rpm
Token rpm
Feature activation+0.000
files
Token files
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
an
Token an
Feature activation+2.814
older
Token older
Feature activation+0.309
version
Token version
Feature activation+5.106
is
Token is
Feature activation+3.350
already
Token already
Feature activation+1.647
installed
Token installed
Feature activation+0.000
outright
Token outright
Feature activation+0.000
lies
Token lies
Feature activation+0.000
.
Token.
Feature activation+0.000
Of
Token Of
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+2.757
should
Token should
Feature activation+4.347
highlight
Token highlight
Feature activation+4.498
,
Token,
Feature activation+0.485
Treasury
Token Treasury
Feature activation+0.000
Secretary
Token Secretary
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
an
Token an
Feature activation+2.814
older
Token older
Feature activation+0.309
version
Token version
Feature activation+5.106
is
Token is
Feature activation+3.350
already
Token already
Feature activation+1.647
installed
Token installed
Feature activation+0.000
will
Token will
Feature activation+0.000
be
Token be
Feature activation+0.000
installed
Token installed
Feature activation+0.000
which
Token which
Feature activation+0.000
around
Token around
Feature activation+1.637
$
Token $
Feature activation+2.112
8
Token8
Feature activation+0.391
,
Token,
Feature activation+0.000
000
Token000
Feature activation+3.016
goes
Token goes
Feature activation+0.000
to
Token to
Feature activation+0.010
the
Token the
Feature activation+0.000
surrogate
Token surrogate
Feature activation+0.000
mother
Token mother
Feature activation+0.000

INTERVAL 1.726 - 2.589
CONTAINS 0.001%

which
Token which
Feature activation+0.000
I
Token I
Feature activation+5.213
am
Token am
Feature activation+6.344
at
Token at
Feature activation+2.998
the
Token the
Feature activation+3.760
tail
Token tail
Feature activation+1.956
end
Token end
Feature activation+6.364
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
complete
Token complete
Feature activation+0.000
the
Token the
Feature activation+0.000
unit
Token unit
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
his
Token his
Feature activation+2.449
or
Token or
Feature activation+0.000
her
Token her
Feature activation+3.089
purchase
Token purchase
Feature activation+2.398
is
Token is
Feature activation+3.663
a
Token a
Feature activation+2.966
28
Token 28
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
24
Token 24
Feature activation+2.002
inches
Token inches
Feature activation+1.054
is
Token is
Feature activation+0.821
reserved
Token reserved
Feature activation+0.706
on
Token on
Feature activation+0.212
the
Token the
Feature activation+0.000
President
Token President
Feature activation+0.000
Obama
Token Obama
Feature activation+0.000
,
Token,
Feature activation+0.000
about
Token about
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+2.160
writes
Token writes
Feature activation+2.462
,
Token,
Feature activation+0.000
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.000
[
Token[
Feature activation+0.216
target
Token target
Feature activation+0.000
players
Token players
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
nearly
Token nearly
Feature activation+1.938
all
Token all
Feature activation+1.925
come
Token come
Feature activation+0.000
from
Token from
Feature activation+0.000
the
Token the
Feature activation+0.000
domestic
Token domestic
Feature activation+0.000
leagues
Token leagues
Feature activation+0.000

INTERVAL 0.863 - 1.726
CONTAINS 0.002%

Treasury
Token Treasury
Feature activation+0.000
Secretary
Token Secretary
Feature activation+0.000
Mn
Token Mn
Feature activation+0.000
uchin
Tokenuchin
Feature activation+0.000
,
Token,
Feature activation+0.000
one
Token one
Feature activation+1.173
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
asserted
Token asserted
Feature activation+0.000
adults
Token adults
Feature activation+0.000
,
Token,
Feature activation+0.000
football
Token football
Feature activation+0.000
in
Token in
Feature activation+0.000
overtime
Token overtime
Feature activation+0.000
,
Token,
Feature activation+0.000
22
Token 22
Feature activation+0.000
of
Token of
Feature activation+0.937
them
Token them
Feature activation+0.000
ended
Token ended
Feature activation+0.000
in
Token in
Feature activation+0.000
a
Token a
Feature activation+0.000
score
Token score
Feature activation+0.000
Bak
Token Bak
Feature activation+0.000
shi
Tokenshi
Feature activation+0.000
,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
around
Token around
Feature activation+1.637
$
Token $
Feature activation+2.112
8
Token8
Feature activation+0.391
,
Token,
Feature activation+0.000
000
Token000
Feature activation+3.016
goes
Token goes
Feature activation+0.000
GG
TokenGG
Feature activation+0.000
are
Token are
Feature activation+5.696
the
Token the
Feature activation+5.726
P
Token P
Feature activation+0.000
AM
TokenAM
Feature activation+0.139
sequence
Token sequence
Feature activation+1.298
(
Token (
Feature activation+0.000
Fig
TokenFig
Feature activation+0.000
.
Token.
Feature activation+0.000
1
Token 1
Feature activation+0.000
).
Token ).
Feature activation+0.000
,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+0.000
fully
Token fully
Feature activation+0.000
complied
Token complied
Feature activation+1.325
with
Token with
Feature activation+0.000
just
Token just
Feature activation+0.000
over
Token over
Feature activation+0.000
half
Token half
Feature activation+1.552
;
Token;
Feature activation+0.000

INTERVAL 0.000 - 0.863
CONTAINS 99.992%

blue
Token blue
Feature activation+0.000
berry
Tokenberry
Feature activation+0.000
farm
Token farm
Feature activation+0.000
had
Token had
Feature activation+0.000
been
Token been
Feature activation+0.000
deported
Token deported
Feature activation+0.000
six
Token six
Feature activation+0.000
times
Token times
Feature activation+0.000
,
Token,
Feature activation+0.000
most
Token most
Feature activation+0.000
recently
Token recently
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Even
TokenEven
Feature activation+0.000
further
Token further
Feature activation+0.000
back
Token back
Feature activation+0.000
,
Token,
Feature activation+0.000
in
Token in
Feature activation+0.000
1897
Token 1897
Feature activation+0.000
,
Token,
Feature activation+0.000
Scientific
Token Scientific
Feature activation+0.000
American
Token American
Feature activation+0.000
noted
Token noted
Feature activation+0.000
)
Token)
Feature activation+0.000
and
Token and
Feature activation+0.000
left
Token left
Feature activation+0.000
fielder
Token fielder
Feature activation+0.000
Shane
Token Shane
Feature activation+0.000
Peterson
Token Peterson
Feature activation+0.000
collide
Token collide
Feature activation+0.000
while
Token while
Feature activation+0.000
trying
Token trying
Feature activation+0.000
to
Token to
Feature activation+0.000
catch
Token catch
Feature activation+0.000
alternative
Token alternative
Feature activation+0.000
option
Token option
Feature activation+0.000
was
Token was
Feature activation+0.000
to
Token to
Feature activation+0.000
place
Token place
Feature activation+0.000
the
Token the
Feature activation+0.000
pump
Token pump
Feature activation+0.000
inside
Token inside
Feature activation+0.000
the
Token the
Feature activation+0.000
locker
Token locker
Feature activation+0.000
and
Token and
Feature activation+0.000
media
Token media
Feature activation+0.000
player
Token player
Feature activation+0.000
.
Token.
Feature activation+0.000
iTunes
Token iTunes
Feature activation+0.000
has
Token has
Feature activation+0.000
been
Token been
Feature activation+0.000
designed
Token designed
Feature activation+0.000
to
Token to
Feature activation+0.000
help
Token help
Feature activation+0.000
users
Token users
Feature activation+0.000
subscribe
Token subscribe
Feature activation+0.000

BOTTOM ACTIVATIONS
MIN = 0.000

atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
Development
Token Development
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000

Top feature 1 in H7.9: (feature 25388)

TOP ACTIVATIONS
MAX = 7.584

State
Token State
Feature activation+0.000
of
Token of
Feature activation+0.000
Palestine
Token Palestine
Feature activation+0.383
is
Token is
Feature activation+6.956
a
Token a
Feature activation+4.519
party
Token party
Feature activation+7.584
âĢĵ
Token âĢĵ
Feature activation+0.000
and
Token and
Feature activation+0.000
not
Token not
Feature activation+0.000
carry
Token carry
Feature activation+0.049
out
Token out
Feature activation+0.000
and
Token and
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.096
dog
Token dog
Feature activation+3.345
responds
Token responds
Feature activation+7.264
appropriately
Token appropriately
Feature activation+3.938
.
Token.
Feature activation+0.000
For
Token For
Feature activation+0.000
example
Token example
Feature activation+0.000
my
Token my
Feature activation+0.000
figure
Token figure
Feature activation+0.000
sk
Token sk
Feature activation+0.000
ater
Tokenater
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
she
Token she
Feature activation+6.985
finished
Token finished
Feature activation+4.515
in
Token in
Feature activation+0.155
third
Token third
Feature activation+0.000
place
Token place
Feature activation+2.679
at
Token at
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.577
State
Token State
Feature activation+0.000
of
Token of
Feature activation+0.000
Palestine
Token Palestine
Feature activation+0.383
is
Token is
Feature activation+6.956
a
Token a
Feature activation+4.519
party
Token party
Feature activation+7.584
âĢĵ
Token âĢĵ
Feature activation+0.000
and
Token and
Feature activation+0.000
not
Token not
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
only
Token only
Feature activation+0.000
he
Token he
Feature activation+4.625
had
Token had
Feature activation+5.493
access
Token access
Feature activation+6.538
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
part
Token part
Feature activation+0.000
of
Token of
Feature activation+0.000
his
Token his
Feature activation+0.000
length
Token length
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
you
Token you
Feature activation+0.000
floor
Token floor
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
safe
Token safe
Feature activation+0.000
had
Token had
Feature activation+6.220
been
Token been
Feature activation+5.467
delivered
Token delivered
Feature activation+0.168
were
Token were
Feature activation+0.000
sent
Token sent
Feature activation+0.000
home
Token home
Feature activation+0.000
loved
Token loved
Feature activation+0.000
ones
Token ones
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+5.415
offer
Token offer
Feature activation+5.604
our
Token our
Feature activation+2.119
sincere
Token sincere
Feature activation+0.857
condolences
Token condolences
Feature activation+1.943
and
Token and
Feature activation+0.000
assure
Token assure
Feature activation+0.000
exchanges
Token exchanges
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
only
Token only
Feature activation+0.000
he
Token he
Feature activation+4.625
had
Token had
Feature activation+5.493
access
Token access
Feature activation+6.538
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
part
Token part
Feature activation+0.000
of
Token of
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
safe
Token safe
Feature activation+0.000
had
Token had
Feature activation+6.220
been
Token been
Feature activation+5.467
delivered
Token delivered
Feature activation+0.168
were
Token were
Feature activation+0.000
sent
Token sent
Feature activation+0.000
home
Token home
Feature activation+0.000
early
Token early
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
the
Token the
Feature activation+0.000
Knicks
Token Knicks
Feature activation+2.486
possibly
Token possibly
Feature activation+4.141
had
Token had
Feature activation+5.448
a
Token a
Feature activation+3.392
silent
Token silent
Feature activation+0.000
agreement
Token agreement
Feature activation+5.316
with
Token with
Feature activation+5.105
to
Token to
Feature activation+0.000
and
Token and
Feature activation+0.000
loved
Token loved
Feature activation+0.000
ones
Token ones
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+5.415
offer
Token offer
Feature activation+5.604
our
Token our
Feature activation+2.119
sincere
Token sincere
Feature activation+0.857
condolences
Token condolences
Feature activation+1.943
and
Token and
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
you
Token you
Feature activation+0.000
tell
Token tell
Feature activation+0.000
Knicks
Token Knicks
Feature activation+2.486
possibly
Token possibly
Feature activation+4.141
had
Token had
Feature activation+5.448
a
Token a
Feature activation+3.392
silent
Token silent
Feature activation+0.000
agreement
Token agreement
Feature activation+5.316
with
Token with
Feature activation+5.105
to
Token to
Feature activation+0.000
re
Token re
Feature activation+0.000
-
Token-
Feature activation+0.000
sign
Tokensign
Feature activation+1.076
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
you
Token you
Feature activation+0.000
tell
Token tell
Feature activation+0.000
people
Token people
Feature activation+0.000
possibly
Token possibly
Feature activation+4.141
had
Token had
Feature activation+5.448
a
Token a
Feature activation+3.392
silent
Token silent
Feature activation+0.000
agreement
Token agreement
Feature activation+5.316
with
Token with
Feature activation+5.105
to
Token to
Feature activation+0.000
re
Token re
Feature activation+0.000
-
Token-
Feature activation+0.000
sign
Tokensign
Feature activation+1.076
on
Token on
Feature activation+0.000
insane
Token insane
Feature activation+0.000
length
Token length
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
bitcoin
Token bitcoin
Feature activation+0.000
exchanges
Token exchanges
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
only
Token only
Feature activation+0.000
he
Token he
Feature activation+4.625
had
Token had
Feature activation+5.493
access
Token access
Feature activation+6.538
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
part
Token part
Feature activation+0.000
the
Token the
Feature activation+0.577
State
Token State
Feature activation+0.000
of
Token of
Feature activation+0.000
Palestine
Token Palestine
Feature activation+0.383
is
Token is
Feature activation+6.956
a
Token a
Feature activation+4.519
party
Token party
Feature activation+7.584
âĢĵ
Token âĢĵ
Feature activation+0.000
and
Token and
Feature activation+0.000
not
Token not
Feature activation+0.000
carry
Token carry
Feature activation+0.049
sk
Token sk
Feature activation+0.000
ater
Tokenater
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
she
Token she
Feature activation+6.985
finished
Token finished
Feature activation+4.515
in
Token in
Feature activation+0.155
third
Token third
Feature activation+0.000
place
Token place
Feature activation+2.679
at
Token at
Feature activation+0.000
a
Token a
Feature activation+0.000

Top DFA by src position
MAX = 8.833

and
Token and
Feature activation-0.004
Top resid features:
Political
Token Political
Feature activation-0.001
Top resid features:
Rights
Token Rights
Feature activation-0.007
Top resid features:
,
Token,
Feature activation-0.003
Top resid features:
to
Token to
Feature activation+0.013
Top resid features:
which
Token which
Feature activation+8.833
Top resid features:
the
Token the
Feature activation+0.077
Top resid features:
State
Token State
Feature activation-0.023
Top resid features:
of
Token of
Feature activation-0.022
Top resid features:
Palestine
Token Palestine
Feature activation-0.070
Top resid features:
is
Token is
Feature activation+0.165
Top resid features:
that
Token that
Feature activation+0.001
Top resid features:
they
Token they
Feature activation+0.015
Top resid features:
use
Token use
Feature activation+0.008
Top resid features:
and
Token and
Feature activation+0.007
Top resid features:
to
Token to
Feature activation+0.077
Top resid features:
which
Token which
Feature activation+8.355
Top resid features:
the
Token the
Feature activation+0.108
Top resid features:
dog
Token dog
Feature activation+0.033
Top resid features:
responds
Token responds
Feature activation+0.186
Top resid features:
appropriately
Token appropriately
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
competitive
Token competitive
Feature activation-0.068
Top resid features:
figure
Token figure
Feature activation+0.033
Top resid features:
sk
Token sk
Feature activation+0.017
Top resid features:
ater
Tokenater
Feature activation+0.012
Top resid features:
to
Token to
Feature activation+0.110
Top resid features:
which
Token which
Feature activation+8.434
Top resid features:
she
Token she
Feature activation+0.092
Top resid features:
finished
Token finished
Feature activation+0.000
Top resid features:
in
Token in
Feature activation+0.000
Top resid features:
third
Token third
Feature activation+0.000
Top resid features:
place
Token place
Feature activation+0.000
Top resid features:
and
Token and
Feature activation-0.009
Top resid features:
Political
Token Political
Feature activation-0.005
Top resid features:
Rights
Token Rights
Feature activation-0.014
Top resid features:
,
Token,
Feature activation-0.004
Top resid features:
to
Token to
Feature activation+0.010
Top resid features:
which
Token which
Feature activation+8.564
Top resid features:
the
Token the
Feature activation+0.092
Top resid features:
State
Token State
Feature activation-0.137
Top resid features:
of
Token of
Feature activation-0.148
Top resid features:
Palestine
Token Palestine
Feature activation+0.032
Top resid features:
is
Token is
Feature activation+0.040
Top resid features:
at
Token at
Feature activation+0.018
Top resid features:
other
Token other
Feature activation-0.003
Top resid features:
bitcoin
Token bitcoin
Feature activation-0.003
Top resid features:
exchanges
Token exchanges
Feature activation-0.015
Top resid features:
to
Token to
Feature activation+0.045
Top resid features:
which
Token which
Feature activation+7.715
Top resid features:
only
Token only
Feature activation+0.056
Top resid features:
he
Token he
Feature activation+0.131
Top resid features:
had
Token had
Feature activation+0.194
Top resid features:
access
Token access
Feature activation+0.086
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
is
Token is
Feature activation-0.004
Top resid features:
the
Token the
Feature activation-0.033
Top resid features:
insane
Token insane
Feature activation-0.018
Top resid features:
length
Token length
Feature activation-0.012
Top resid features:
to
Token to
Feature activation+0.142
Top resid features:
which
Token which
Feature activation+7.767
Top resid features:
political
Token political
Feature activation+0.004
Top resid features:
correctness
Token correctness
Feature activation+0.155
Top resid features:
has
Token has
Feature activation+0.043
Top resid features:
taken
Token taken
Feature activation+0.000
Top resid features:
us
Token us
Feature activation+0.000
Top resid features:
working
Token working
Feature activation+0.020
Top resid features:
on
Token on
Feature activation-0.056
Top resid features:
the
Token the
Feature activation-0.039
Top resid features:
floor
Token floor
Feature activation-0.021
Top resid features:
to
Token to
Feature activation+0.044
Top resid features:
which
Token which
Feature activation+7.936
Top resid features:
the
Token the
Feature activation+0.111
Top resid features:
safe
Token safe
Feature activation-0.173
Top resid features:
had
Token had
Feature activation+0.083
Top resid features:
been
Token been
Feature activation+0.000
Top resid features:
delivered
Token delivered
Feature activation+0.000
Top resid features:
family
Token family
Feature activation+0.026
Top resid features:
and
Token and
Feature activation-0.002
Top resid features:
loved
Token loved
Feature activation-0.008
Top resid features:
ones
Token ones
Feature activation-0.034
Top resid features:
to
Token to
Feature activation+0.035
Top resid features:
whom
Token whom
Feature activation+6.329
Top resid features:
we
Token we
Feature activation+0.983
Top resid features:
offer
Token offer
Feature activation+0.082
Top resid features:
our
Token our
Feature activation+0.000
Top resid features:
sincere
Token sincere
Feature activation+0.000
Top resid features:
condolences
Token condolences
Feature activation+0.000
Top resid features:
at
Token at
Feature activation+0.023
Top resid features:
other
Token other
Feature activation-0.019
Top resid features:
bitcoin
Token bitcoin
Feature activation+0.003
Top resid features:
exchanges
Token exchanges
Feature activation-0.011
Top resid features:
to
Token to
Feature activation+0.023
Top resid features:
which
Token which
Feature activation+6.622
Top resid features:
only
Token only
Feature activation+0.182
Top resid features:
he
Token he
Feature activation+0.504
Top resid features:
had
Token had
Feature activation+0.020
Top resid features:
access
Token access
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
working
Token working
Feature activation+0.014
Top resid features:
on
Token on
Feature activation-0.042
Top resid features:
the
Token the
Feature activation-0.033
Top resid features:
floor
Token floor
Feature activation-0.021
Top resid features:
to
Token to
Feature activation+0.017
Top resid features:
which
Token which
Feature activation+7.191
Top resid features:
the
Token the
Feature activation+0.021
Top resid features:
safe
Token safe
Feature activation-0.224
Top resid features:
had
Token had
Feature activation+0.094
Top resid features:
been
Token been
Feature activation+0.037
Top resid features:
delivered
Token delivered
Feature activation+0.000
Top resid features:
a
Token a
Feature activation-0.012
Top resid features:
C
Token C
Feature activation-0.002
Top resid features:
AA
TokenAA
Feature activation+0.020
Top resid features:
client
Token client
Feature activation+0.039
Top resid features:
to
Token to
Feature activation+0.074
Top resid features:
whom
Token whom
Feature activation+6.411
Top resid features:
the
Token the
Feature activation+0.088
Top resid features:
Knicks
Token Knicks
Feature activation+0.021
Top resid features:
possibly
Token possibly
Feature activation+0.290
Top resid features:
had
Token had
Feature activation+0.056
Top resid features:
a
Token a
Feature activation+0.000
Top resid features:
family
Token family
Feature activation+0.040
Top resid features:
and
Token and
Feature activation-0.031
Top resid features:
loved
Token loved
Feature activation-0.155
Top resid features:
ones
Token ones
Feature activation+0.008
Top resid features:
to
Token to
Feature activation+0.077
Top resid features:
whom
Token whom
Feature activation+7.255
Top resid features:
we
Token we
Feature activation+0.075
Top resid features:
offer
Token offer
Feature activation+0.000
Top resid features:
our
Token our
Feature activation+0.000
Top resid features:
sincere
Token sincere
Feature activation+0.000
Top resid features:
condolences
Token condolences
Feature activation+0.000
Top resid features:
is
Token is
Feature activation-0.009
Top resid features:
the
Token the
Feature activation-0.055
Top resid features:
insane
Token insane
Feature activation-0.033
Top resid features:
length
Token length
Feature activation+0.012
Top resid features:
to
Token to
Feature activation+0.171
Top resid features:
which
Token which
Feature activation+6.661
Top resid features:
political
Token political
Feature activation-0.010
Top resid features:
correctness
Token correctness
Feature activation-0.023
Top resid features:
has
Token has
Feature activation+0.118
Top resid features:
taken
Token taken
Feature activation+0.153
Top resid features:
us
Token us
Feature activation+0.000
Top resid features:
a
Token a
Feature activation+0.013
Top resid features:
C
Token C
Feature activation+0.005
Top resid features:
AA
TokenAA
Feature activation+0.000
Top resid features:
client
Token client
Feature activation+0.010
Top resid features:
to
Token to
Feature activation+0.088
Top resid features:
whom
Token whom
Feature activation+6.659
Top resid features:
the
Token the
Feature activation+0.075
Top resid features:
Knicks
Token Knicks
Feature activation+0.013
Top resid features:
possibly
Token possibly
Feature activation+0.102
Top resid features:
had
Token had
Feature activation+0.041
Top resid features:
a
Token a
Feature activation-0.035
Top resid features:
is
Token is
Feature activation-0.013
Top resid features:
the
Token the
Feature activation-0.027
Top resid features:
insane
Token insane
Feature activation-0.043
Top resid features:
length
Token length
Feature activation+0.028
Top resid features:
to
Token to
Feature activation+0.153
Top resid features:
which
Token which
Feature activation+6.480
Top resid features:
political
Token political
Feature activation-0.008
Top resid features:
correctness
Token correctness
Feature activation-0.010
Top resid features:
has
Token has
Feature activation+0.065
Top resid features:
taken
Token taken
Feature activation+0.230
Top resid features:
us
Token us
Feature activation+0.009
Top resid features:
a
Token a
Feature activation+0.010
Top resid features:
C
Token C
Feature activation+0.004
Top resid features:
AA
TokenAA
Feature activation+0.039
Top resid features:
client
Token client
Feature activation+0.032
Top resid features:
to
Token to
Feature activation+0.061
Top resid features:
whom
Token whom
Feature activation+6.382
Top resid features:
the
Token the
Feature activation+0.070
Top resid features:
Knicks
Token Knicks
Feature activation+0.009
Top resid features:
possibly
Token possibly
Feature activation+0.073
Top resid features:
had
Token had
Feature activation+0.014
Top resid features:
a
Token a
Feature activation-0.007
Top resid features:
is
Token is
Feature activation-0.008
Top resid features:
the
Token the
Feature activation-0.034
Top resid features:
insane
Token insane
Feature activation-0.021
Top resid features:
length
Token length
Feature activation-0.002
Top resid features:
to
Token to
Feature activation+0.099
Top resid features:
which
Token which
Feature activation+6.162
Top resid features:
political
Token political
Feature activation-0.010
Top resid features:
correctness
Token correctness
Feature activation+0.014
Top resid features:
has
Token has
Feature activation+0.000
Top resid features:
taken
Token taken
Feature activation+0.000
Top resid features:
us
Token us
Feature activation+0.000
Top resid features:
at
Token at
Feature activation+0.028
Top resid features:
other
Token other
Feature activation-0.015
Top resid features:
bitcoin
Token bitcoin
Feature activation-0.008
Top resid features:
exchanges
Token exchanges
Feature activation-0.029
Top resid features:
to
Token to
Feature activation+0.056
Top resid features:
which
Token which
Feature activation+6.090
Top resid features:
only
Token only
Feature activation+0.185
Top resid features:
he
Token he
Feature activation+0.012
Top resid features:
had
Token had
Feature activation+0.000
Top resid features:
access
Token access
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
and
Token and
Feature activation-0.005
Top resid features:
Political
Token Political
Feature activation-0.007
Top resid features:
Rights
Token Rights
Feature activation-0.009
Top resid features:
,
Token,
Feature activation-0.003
Top resid features:
to
Token to
Feature activation+0.007
Top resid features:
which
Token which
Feature activation+6.207
Top resid features:
the
Token the
Feature activation+0.063
Top resid features:
State
Token State
Feature activation-0.110
Top resid features:
of
Token of
Feature activation-0.217
Top resid features:
Palestine
Token Palestine
Feature activation-0.105
Top resid features:
is
Token is
Feature activation+0.120
Top resid features:
competitive
Token competitive
Feature activation-0.067
Top resid features:
figure
Token figure
Feature activation+0.012
Top resid features:
sk
Token sk
Feature activation+0.037
Top resid features:
ater
Tokenater
Feature activation+0.022
Top resid features:
to
Token to
Feature activation+0.073
Top resid features:
which
Token which
Feature activation+5.707
Top resid features:
she
Token she
Feature activation+0.238
Top resid features:
finished
Token finished
Feature activation+0.053
Top resid features:
in
Token in
Feature activation+0.000
Top resid features:
third
Token third
Feature activation+0.000
Top resid features:
place
Token place
Feature activation+0.000
Top resid features:

Decoder Weights Distribution

Head 0: 0.05

Head 1: 0.02

Head 2: 0.01

Head 3: 0.06

Head 4: 0.07

Head 5: 0.04

Head 6: 0.03

Head 7: 0.01

Head 8: 0.16

Head 9: 0.53

Head 10: 0.01

Head 11: 0.02

Positive logits

oyal2.38

aspire2.26

belongs2.23

refers1.97

reply1.95

ributes1.94

malink1.90

alluded1.90

relates1.88

udes1.87

ptr1.86

umb1.83

homage1.82

uding1.81

pointers1.79

ember1.78

cling1.77

applies1.75

Alc1.74

attachment1.73

Negative logits

yout-2.01

raided-1.98

Solitaire-1.95

artifacts-1.93

FANTASY-1.92

Strike-1.92

iphany-1.92

Reviewer-1.91

arettes-1.90

yrinth-1.85

BAT-1.84

uala-1.83

STOP-1.82

uto-1.80

stocks-1.79

maxwell-1.78

spotting-1.74

nir-1.74

Benny-1.74

MARK-1.74

INTERVAL 6.825 - 7.584
CONTAINS 0.000%

and
Token and
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.096
dog
Token dog
Feature activation+3.345
responds
Token responds
Feature activation+7.264
appropriately
Token appropriately
Feature activation+3.938
.
Token.
Feature activation+0.000
For
Token For
Feature activation+0.000
example
Token example
Feature activation+0.000
my
Token my
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.577
State
Token State
Feature activation+0.000
of
Token of
Feature activation+0.000
Palestine
Token Palestine
Feature activation+0.383
is
Token is
Feature activation+6.956
a
Token a
Feature activation+4.519
party
Token party
Feature activation+7.584
âĢĵ
Token âĢĵ
Feature activation+0.000
and
Token and
Feature activation+0.000
not
Token not
Feature activation+0.000
State
Token State
Feature activation+0.000
of
Token of
Feature activation+0.000
Palestine
Token Palestine
Feature activation+0.383
is
Token is
Feature activation+6.956
a
Token a
Feature activation+4.519
party
Token party
Feature activation+7.584
âĢĵ
Token âĢĵ
Feature activation+0.000
and
Token and
Feature activation+0.000
not
Token not
Feature activation+0.000
carry
Token carry
Feature activation+0.049
out
Token out
Feature activation+0.000
figure
Token figure
Feature activation+0.000
sk
Token sk
Feature activation+0.000
ater
Tokenater
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
she
Token she
Feature activation+6.985
finished
Token finished
Feature activation+4.515
in
Token in
Feature activation+0.155
third
Token third
Feature activation+0.000
place
Token place
Feature activation+2.679
at
Token at
Feature activation+0.000

INTERVAL 6.067 - 6.825
CONTAINS 0.000%

to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
only
Token only
Feature activation+0.000
he
Token he
Feature activation+4.625
had
Token had
Feature activation+5.493
access
Token access
Feature activation+6.538
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
part
Token part
Feature activation+0.000
of
Token of
Feature activation+0.000
his
Token his
Feature activation+0.000
floor
Token floor
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
safe
Token safe
Feature activation+0.000
had
Token had
Feature activation+6.220
been
Token been
Feature activation+5.467
delivered
Token delivered
Feature activation+0.168
were
Token were
Feature activation+0.000
sent
Token sent
Feature activation+0.000
home
Token home
Feature activation+0.000
length
Token length
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
you
Token you
Feature activation+0.000

INTERVAL 5.309 - 6.067
CONTAINS 0.001%

to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
the
Token the
Feature activation+0.000
Knicks
Token Knicks
Feature activation+2.486
possibly
Token possibly
Feature activation+4.141
had
Token had
Feature activation+5.448
a
Token a
Feature activation+3.392
silent
Token silent
Feature activation+0.000
agreement
Token agreement
Feature activation+5.316
with
Token with
Feature activation+5.105
to
Token to
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
you
Token you
Feature activation+0.000
tell
Token tell
Feature activation+0.000
and
Token and
Feature activation+0.000
loved
Token loved
Feature activation+0.000
ones
Token ones
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+5.415
offer
Token offer
Feature activation+5.604
our
Token our
Feature activation+2.119
sincere
Token sincere
Feature activation+0.857
condolences
Token condolences
Feature activation+1.943
and
Token and
Feature activation+0.000
loved
Token loved
Feature activation+0.000
ones
Token ones
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+5.415
offer
Token offer
Feature activation+5.604
our
Token our
Feature activation+2.119
sincere
Token sincere
Feature activation+0.857
condolences
Token condolences
Feature activation+1.943
and
Token and
Feature activation+0.000
assure
Token assure
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
safe
Token safe
Feature activation+0.000
had
Token had
Feature activation+6.220
been
Token been
Feature activation+5.467
delivered
Token delivered
Feature activation+0.168
were
Token were
Feature activation+0.000
sent
Token sent
Feature activation+0.000
home
Token home
Feature activation+0.000
early
Token early
Feature activation+0.000

INTERVAL 4.550 - 5.309
CONTAINS 0.000%

bitcoin
Token bitcoin
Feature activation+0.000
exchanges
Token exchanges
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
only
Token only
Feature activation+0.000
he
Token he
Feature activation+4.625
had
Token had
Feature activation+5.493
access
Token access
Feature activation+6.538
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
part
Token part
Feature activation+0.000
possibly
Token possibly
Feature activation+4.141
had
Token had
Feature activation+5.448
a
Token a
Feature activation+3.392
silent
Token silent
Feature activation+0.000
agreement
Token agreement
Feature activation+5.316
with
Token with
Feature activation+5.105
to
Token to
Feature activation+0.000
re
Token re
Feature activation+0.000
-
Token-
Feature activation+0.000
sign
Tokensign
Feature activation+1.076
on
Token on
Feature activation+0.000
insane
Token insane
Feature activation+0.000
length
Token length
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
which
Token which
Feature activation+0.000
political
Token political
Feature activation+0.000
correctness
Token correctness
Feature activation+4.795
has
Token has
Feature activation+6.468
taken
Token taken
Feature activation+5.345
us
Token us
Feature activation+5.251
:
Token:
Feature activation+0.000
if
Token if
Feature activation+0.000
you
Token you
Feature activation+0.000
tell
Token tell
Feature activation+0.000
people
Token people
Feature activation+0.000

INTERVAL 3.792 - 4.550
CONTAINS 0.001%

client
Token client
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
the
Token the
Feature activation+0.000
Knicks
Token Knicks
Feature activation+2.486
possibly
Token possibly
Feature activation+4.141
had
Token had
Feature activation+5.448
a
Token a
Feature activation+3.392
silent
Token silent
Feature activation+0.000
agreement
Token agreement
Feature activation+5.316
with
Token with
Feature activation+5.105
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.096
dog
Token dog
Feature activation+3.345
responds
Token responds
Feature activation+7.264
appropriately
Token appropriately
Feature activation+3.938
.
Token.
Feature activation+0.000
For
Token For
Feature activation+0.000
example
Token example
Feature activation+0.000
my
Token my
Feature activation+0.000
dogs
Token dogs
Feature activation+0.000
sk
Token sk
Feature activation+0.000
ater
Tokenater
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
she
Token she
Feature activation+6.985
finished
Token finished
Feature activation+4.515
in
Token in
Feature activation+0.155
third
Token third
Feature activation+0.000
place
Token place
Feature activation+2.679
at
Token at
Feature activation+0.000
a
Token a
Feature activation+0.000
052
Token052
Feature activation+0.000
),
Token),
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
people
Token people
Feature activation+3.430
travelled
Token travelled
Feature activation+4.187
from
Token from
Feature activation+0.099
across
Token across
Feature activation+0.000
the
Token the
Feature activation+0.000
east
Token east
Feature activation+0.000
coast
Token coast
Feature activation+0.000
ino
Tokenino
Feature activation+0.000
,
Token,
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.212
editor
Token editor
Feature activation+4.333
responded
Token responded
Feature activation+2.363
,
Token,
Feature activation+0.000
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.000
The
TokenThe
Feature activation+0.000

INTERVAL 3.034 - 3.792
CONTAINS 0.001%

know
Token know
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
you
Token you
Feature activation+2.931
're
Token're
Feature activation+2.835
referring
Token referring
Feature activation+3.231
.
Token.
Feature activation+0.000
S
Token S
Feature activation+0.000
ibling
Tokenibling
Feature activation+0.000
is
Token is
Feature activation+0.000
more
Token more
Feature activation+0.000
for
Token for
Feature activation+0.000
the
Token the
Feature activation+0.000
generations
Token generations
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+3.152
will
Token will
Feature activation+3.516
pass
Token pass
Feature activation+0.499
them
Token them
Feature activation+0.000
on
Token on
Feature activation+0.000
,
Token,
Feature activation+0.000
the
Token the
Feature activation+0.000
generations
Token generations
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+3.152
will
Token will
Feature activation+3.516
pass
Token pass
Feature activation+0.499
them
Token them
Feature activation+0.000
on
Token on
Feature activation+0.000
,
Token,
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
use
Token use
Feature activation+0.000
and
Token and
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.096
dog
Token dog
Feature activation+3.345
responds
Token responds
Feature activation+7.264
appropriately
Token appropriately
Feature activation+3.938
.
Token.
Feature activation+0.000
For
Token For
Feature activation+0.000
example
Token example
Feature activation+0.000
whom
Token whom
Feature activation+0.000
the
Token the
Feature activation+0.000
Knicks
Token Knicks
Feature activation+2.486
possibly
Token possibly
Feature activation+4.141
had
Token had
Feature activation+5.448
a
Token a
Feature activation+3.392
silent
Token silent
Feature activation+0.000
agreement
Token agreement
Feature activation+5.316
with
Token with
Feature activation+5.105
to
Token to
Feature activation+0.000
re
Token re
Feature activation+0.000

INTERVAL 2.275 - 3.034
CONTAINS 0.001%

against
Token against
Feature activation+0.000
a
Token a
Feature activation+0.000
society
Token society
Feature activation+0.000
into
Token into
Feature activation+0.000
which
Token which
Feature activation+0.000
they
Token they
Feature activation+2.518
were
Token were
Feature activation+2.375
born
Token born
Feature activation+1.755
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+2.327
are
Token are
Feature activation+2.514
all
Token all
Feature activation+2.644
so
Token so
Feature activation+2.398
very
Token very
Feature activation+2.293
,
Token,
Feature activation+0.000
very
Token very
Feature activation+1.695
proud
Token proud
Feature activation+2.123
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
to
Token to
Feature activation+0.000
know
Token know
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
you
Token you
Feature activation+2.931
're
Token're
Feature activation+2.835
referring
Token referring
Feature activation+3.231
.
Token.
Feature activation+0.000
S
Token S
Feature activation+0.000
ibling
Tokenibling
Feature activation+0.000
is
Token is
Feature activation+0.000
country
Token country
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+2.327
are
Token are
Feature activation+2.514
all
Token all
Feature activation+2.644
so
Token so
Feature activation+2.398
very
Token very
Feature activation+2.293
,
Token,
Feature activation+0.000
very
Token very
Feature activation+1.695
proud
Token proud
Feature activation+2.123
,
Token,
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.212
editor
Token editor
Feature activation+4.333
responded
Token responded
Feature activation+2.363
,
Token,
Feature activation+0.000
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
actual
Token actual
Feature activation+0.000

INTERVAL 1.517 - 2.275
CONTAINS 0.001%

ones
Token ones
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
we
Token we
Feature activation+5.415
offer
Token offer
Feature activation+5.604
our
Token our
Feature activation+2.119
sincere
Token sincere
Feature activation+0.857
condolences
Token condolences
Feature activation+1.943
and
Token and
Feature activation+0.000
assure
Token assure
Feature activation+0.000
of
Token of
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
To
TokenTo
Feature activation+0.000
which
Token which
Feature activation+0.000
I
Token I
Feature activation+1.112
answer
Token answer
Feature activation+2.085
,
Token,
Feature activation+0.000
rather
Token rather
Feature activation+0.000
logically
Token logically
Feature activation+0.000
:
Token:
Feature activation+0.000
with
Token with
Feature activation+0.000
are
Token are
Feature activation+2.514
all
Token all
Feature activation+2.644
so
Token so
Feature activation+2.398
very
Token very
Feature activation+2.293
,
Token,
Feature activation+0.000
very
Token very
Feature activation+1.695
proud
Token proud
Feature activation+2.123
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
you
Token you
Feature activation+0.000
so
Token so
Feature activation+0.000
all
Token all
Feature activation+2.644
so
Token so
Feature activation+2.398
very
Token very
Feature activation+2.293
,
Token,
Feature activation+0.000
very
Token very
Feature activation+1.695
proud
Token proud
Feature activation+2.123
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
you
Token you
Feature activation+0.000
so
Token so
Feature activation+0.000
much
Token much
Feature activation+0.000
are
Tokenare
Feature activation+0.000
,
Token,
Feature activation+0.000
to
Token to
Feature activation+0.000
whom
Token whom
Feature activation+0.000
promises
Token promises
Feature activation+0.000
were
Token were
Feature activation+1.977
made
Token made
Feature activation+1.952
by
Token by
Feature activation+0.000
the
Token the
Feature activation+0.000
Congress
Token Congress
Feature activation+0.000
government
Token government
Feature activation+1.124

INTERVAL 0.758 - 1.517
CONTAINS 0.001%

the
Token the
Feature activation+0.000
public
Token public
Feature activation+0.000
may
Token may
Feature activation+0.086
not
Token not
Feature activation+0.162
be
Token be
Feature activation+1.180
fully
Token fully
Feature activation+0.942
aware
Token aware
Feature activation+0.000
,
Token,
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
Ham
Token Ham
Feature activation+0.000
both
Token both
Feature activation+0.000
of
Token of
Feature activation+0.000
men
Token men
Feature activation+0.000
and
Token and
Feature activation+0.000
States
Token States
Feature activation+0.000
must
Token must
Feature activation+1.460
be
Token be
Feature activation+0.370
referred
Token referred
Feature activation+1.558
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
,
Token,
Feature activation+0.000
or
Token or
Feature activation+0.000
her
Token her
Feature activation+0.637
purchase
Token purchase
Feature activation+0.640
is
Token is
Feature activation+0.959
a
Token a
Feature activation+0.423
part
Token part
Feature activation+1.211
,
Token,
Feature activation+0.000
the
Token the
Feature activation+0.000
probability
Token probability
Feature activation+0.000
of
Token of
Feature activation+0.000
my
Token my
Feature activation+0.000
were
Token were
Feature activation+1.977
made
Token made
Feature activation+1.952
by
Token by
Feature activation+0.000
the
Token the
Feature activation+0.000
Congress
Token Congress
Feature activation+0.000
government
Token government
Feature activation+1.124
which
Token which
Feature activation+0.000
have
Token have
Feature activation+0.000
not
Token not
Feature activation+0.000
been
Token been
Feature activation+0.000
met
Token met
Feature activation+0.000
which
Token which
Feature activation+0.000
his
Token his
Feature activation+0.264
or
Token or
Feature activation+0.000
her
Token her
Feature activation+0.637
purchase
Token purchase
Feature activation+0.640
is
Token is
Feature activation+0.959
a
Token a
Feature activation+0.423
part
Token part
Feature activation+1.211
,
Token,
Feature activation+0.000
the
Token the
Feature activation+0.000
probability
Token probability
Feature activation+0.000

INTERVAL 0.000 - 0.758
CONTAINS 99.993%

and
Token and
Feature activation+0.000
the
Token the
Feature activation+0.000
University
Token University
Feature activation+0.000
of
Token of
Feature activation+0.000
Pug
Token Pug
Feature activation+0.000
et
Tokenet
Feature activation+0.000
Sound
Token Sound
Feature activation+0.000
in
Token in
Feature activation+0.000
Washington
Token Washington
Feature activation+0.000
also
Token also
Feature activation+0.000
told
Token told
Feature activation+0.000
his
Token his
Feature activation+0.000
caucus
Token caucus
Feature activation+0.000
,
Token,
Feature activation+0.000
abandoned
Token abandoned
Feature activation+0.000
ship
Token ship
Feature activation+0.000
.
Token.
Feature activation+0.000
He
Token He
Feature activation+0.000
canceled
Token canceled
Feature activation+0.000
the
Token the
Feature activation+0.000
vote
Token vote
Feature activation+0.000
on
Token on
Feature activation+0.000
.
Token.
Feature activation+0.000
The
Token The
Feature activation+0.000
landscape
Token landscape
Feature activation+0.000
in
Token in
Feature activation+0.000
M
Token M
Feature activation+0.000
ra
Tokenra
Feature activation+0.000
uk
Tokenuk
Feature activation+0.000
U
Token U
Feature activation+0.000
in
Token in
Feature activation+0.000
Burma
Token Burma
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
England
Token England
Feature activation+0.000
,
Token,
Feature activation+0.000
9
Token 9
Feature activation+0.000
,
Token,
Feature activation+0.000
5
Token5
Feature activation+0.000
21
Token21
Feature activation+0.000
in
Token in
Feature activation+0.000
total
Token total
Feature activation+0.000
,
Token,
Feature activation+0.000
decreasing
Token decreasing
Feature activation+0.000
in
Token in
Feature activation+0.000
however
Token however
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
List
TokenList
Feature activation+0.000
of
Token of
Feature activation+0.000
cases
Token cases
Feature activation+0.000
we
Token we
Feature activation+0.000
have
Token have
Feature activation+0.000
evaluated
Token evaluated
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000

BOTTOM ACTIVATIONS
MIN = 0.000

atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
Development
Token Development
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000

Top feature 2 in H7.9: (feature 28587)

TOP ACTIVATIONS
MAX = 5.191

Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
But
TokenBut
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
hammer
Token hammer
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
because
Token because
Feature activation+0.000
you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
hammer
Token hammer
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
that
Token that
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.598
you
Token you
Feature activation+5.191
can
Token can
Feature activation+3.451
do
Token do
Feature activation+3.222
something
Token something
Feature activation+1.058
well
Token well
Feature activation+0.000
,
Token,
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
that
Token that
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.598
you
Token you
Feature activation+5.191
can
Token can
Feature activation+3.451
do
Token do
Feature activation+3.222
something
Token something
Feature activation+1.058
well
Token well
Feature activation+0.000
,
Token,
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
new
Token new
Feature activation+0.575
tools
Token tools
Feature activation+2.712
arrive
Token arrive
Feature activation+1.842
,
Token,
Feature activation+0.257
like
Token like
Feature activation+0.000
3
Token 3
Feature activation+0.000
-
Token-
Feature activation+0.000
you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
hammer
Token hammer
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
t
Tokent
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
new
Token new
Feature activation+0.575
tools
Token tools
Feature activation+2.712
arrive
Token arrive
Feature activation+1.842
,
Token,
Feature activation+0.257
like
Token like
Feature activation+0.000
3
Token 3
Feature activation+0.000
-
Token-
Feature activation+0.000
D
TokenD
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.794
relevant
Token relevant
Feature activation+0.000
information
Token information
Feature activation+1.703
is
Token is
Feature activation+1.707
buried
Token buried
Feature activation+0.000
in
Token in
Feature activation+0.000
a
Token a
Feature activation+0.000
giant
Token giant
Feature activation+0.000
pile
Token pile
Feature activation+0.000
âĢĵ
Token âĢĵ
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.794
relevant
Token relevant
Feature activation+0.000
information
Token information
Feature activation+1.703
is
Token is
Feature activation+1.707
buried
Token buried
Feature activation+0.000
in
Token in
Feature activation+0.000
a
Token a
Feature activation+0.000
giant
Token giant
Feature activation+0.000
see
Token see
Feature activation+0.000
:
Token:
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.538
defense
Token defense
Feature activation+1.681
finally
Token finally
Feature activation+0.680
conceded
Token conceded
Feature activation+0.427
a
Token a
Feature activation+0.000
goal
Token goal
Feature activation+0.000
is
Token is
Feature activation+0.000
become
Token become
Feature activation+0.000
acceptable
Token acceptable
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
God
Token God
Feature activation+1.061
says
Token says
Feature activation+1.641
so
Token so
Feature activation+0.000
?
Token?
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
If
TokenIf
Feature activation+0.000
for
Token for
Feature activation+0.000
Roy
Token Roy
Feature activation+0.000
Moore
Token Moore
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
he
Token he
Feature activation+1.144
has
Token has
Feature activation+0.148
an
Token an
Feature activation+0.000
âĢ
Token âĢ
Feature activation+0.000
ĺ
Tokenĺ
Feature activation+0.000
R
TokenR
Feature activation+0.000
genocide
Token genocide
Feature activation+0.000
become
Token become
Feature activation+0.000
acceptable
Token acceptable
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
God
Token God
Feature activation+1.061
says
Token says
Feature activation+1.641
so
Token so
Feature activation+0.000
?
Token?
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.598
you
Token you
Feature activation+5.191
can
Token can
Feature activation+3.451
do
Token do
Feature activation+3.222
something
Token something
Feature activation+1.058
well
Token well
Feature activation+0.000
,
Token,
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
always
Token always
Feature activation+0.000
just
Token just
Feature activation+0.000
assumed
Token assumed
Feature activation+0.000
because
Token because
Feature activation+0.000
I
Token I
Feature activation+0.476
was
Token was
Feature activation+0.934
part
Token part
Feature activation+0.000
of
Token of
Feature activation+0.607
a
Token a
Feature activation+0.000
certain
Token certain
Feature activation+0.000
family
Token family
Feature activation+0.000
)
Token)
Feature activation+0.000
That
TokenThat
Feature activation+0.000
because
Token because
Feature activation+0.000
I
Token I
Feature activation+0.719
don
Token don
Feature activation+0.471
't
Token't
Feature activation+0.890
consider
Token consider
Feature activation+0.347
any
Token any
Feature activation+0.000
other
Token other
Feature activation+0.000
to
Token to
Feature activation+0.000
actually
Token actually
Feature activation+0.000
at
Token at
Feature activation+0.000
all
Token all
Feature activation+0.000
,
Token,
Feature activation+0.000
simply
Token simply
Feature activation+0.000
because
Token because
Feature activation+0.000
I
Token I
Feature activation+0.842
had
Token had
Feature activation+0.340
other
Token other
Feature activation+0.000
books
Token books
Feature activation+0.000
at
Token at
Feature activation+0.000
hand
Token hand
Feature activation+0.000
found
Token found
Feature activation+0.000
before
Token before
Feature activation+0.000
âĢĵ
Token âĢĵ
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.794
relevant
Token relevant
Feature activation+0.000
information
Token information
Feature activation+1.703
is
Token is
Feature activation+1.707
buried
Token buried
Feature activation+0.000
in
Token in
Feature activation+0.000
just
Token just
Feature activation+0.000
once
Token once
Feature activation+0.000
)
Token)
Feature activation+0.000
That
TokenThat
Feature activation+0.000
because
Token because
Feature activation+0.000
I
Token I
Feature activation+0.719
don
Token don
Feature activation+0.471
't
Token't
Feature activation+0.890
consider
Token consider
Feature activation+0.347
any
Token any
Feature activation+0.000
other
Token other
Feature activation+0.000

Top DFA by src position
MAX = 6.517

.
Token.
Feature activation+0.019
Top resid features:
Ċ
TokenĊ
Feature activation+0.020
Top resid features:
Ċ
TokenĊ
Feature activation-0.004
Top resid features:
But
TokenBut
Feature activation-0.081
Top resid features:
just
Token just
Feature activation+0.054
Top resid features:
because
Token because
Feature activation+6.517
Top resid features:
you
Token you
Feature activation+0.208
Top resid features:
âĢ
TokenâĢ
Feature activation+0.000
Top resid features:
Ļ
TokenĻ
Feature activation+0.000
Top resid features:
ve
Tokenve
Feature activation+0.000
Top resid features:
got
Token got
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.008
Top resid features:
Ċ
TokenĊ
Feature activation+0.010
Top resid features:
Ċ
TokenĊ
Feature activation+0.007
Top resid features:
But
TokenBut
Feature activation-0.023
Top resid features:
just
Token just
Feature activation+0.066
Top resid features:
because
Token because
Feature activation+5.689
Top resid features:
you
Token you
Feature activation+0.767
Top resid features:
âĢ
TokenâĢ
Feature activation+0.011
Top resid features:
Ļ
TokenĻ
Feature activation+0.153
Top resid features:
ve
Tokenve
Feature activation+0.112
Top resid features:
got
Token got
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.017
Top resid features:
Ċ
TokenĊ
Feature activation+0.019
Top resid features:
Ċ
TokenĊ
Feature activation+0.009
Top resid features:
But
TokenBut
Feature activation-0.007
Top resid features:
just
Token just
Feature activation+0.037
Top resid features:
because
Token because
Feature activation+4.307
Top resid features:
you
Token you
Feature activation+0.888
Top resid features:
âĢ
TokenâĢ
Feature activation-0.023
Top resid features:
Ļ
TokenĻ
Feature activation+0.092
Top resid features:
ve
Tokenve
Feature activation+0.201
Top resid features:
got
Token got
Feature activation+0.170
Top resid features:
<|endoftext|>
Token<|endoftext|>
Feature activation+0.989
Top resid features:
that
Token that
Feature activation-0.173
Top resid features:
just
Token just
Feature activation+0.087
Top resid features:
because
Token because
Feature activation+4.459
Top resid features:
you
Token you
Feature activation+0.395
Top resid features:
can
Token can
Feature activation+0.395
Top resid features:
do
Token do
Feature activation+0.000
Top resid features:
something
Token something
Feature activation+0.000
Top resid features:
well
Token well
Feature activation+0.000
Top resid features:
<|endoftext|>
Token<|endoftext|>
Feature activation+0.857
Top resid features:
that
Token that
Feature activation-0.154
Top resid features:
just
Token just
Feature activation+0.107
Top resid features:
because
Token because
Feature activation+3.829
Top resid features:
you
Token you
Feature activation+0.442
Top resid features:
can
Token can
Feature activation+0.794
Top resid features:
do
Token do
Feature activation+0.050
Top resid features:
something
Token something
Feature activation+0.000
Top resid features:
well
Token well
Feature activation+0.000
Top resid features:
No
Token No
Feature activation-0.015
Top resid features:
.
Token.
Feature activation-0.036
Top resid features:
âĢ
TokenâĢ
Feature activation+0.018
Top resid features:
Ŀ
TokenĿ
Feature activation-0.006
Top resid features:
Just
Token Just
Feature activation+0.005
Top resid features:
because
Token because
Feature activation+4.339
Top resid features:
new
Token new
Feature activation+0.340
Top resid features:
tools
Token tools
Feature activation+0.047
Top resid features:
arrive
Token arrive
Feature activation+0.000
Top resid features:
,
Token,
Feature activation+0.000
Top resid features:
like
Token like
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.037
Top resid features:
Ċ
TokenĊ
Feature activation+0.031
Top resid features:
Ċ
TokenĊ
Feature activation+0.020
Top resid features:
But
TokenBut
Feature activation-0.026
Top resid features:
just
Token just
Feature activation+0.090
Top resid features:
because
Token because
Feature activation+2.653
Top resid features:
you
Token you
Feature activation+0.545
Top resid features:
âĢ
TokenâĢ
Feature activation+0.008
Top resid features:
Ļ
TokenĻ
Feature activation+0.022
Top resid features:
ve
Tokenve
Feature activation+0.150
Top resid features:
got
Token got
Feature activation+0.389
Top resid features:
No
Token No
Feature activation-0.011
Top resid features:
.
Token.
Feature activation-0.036
Top resid features:
âĢ
TokenâĢ
Feature activation+0.010
Top resid features:
Ŀ
TokenĿ
Feature activation-0.006
Top resid features:
Just
Token Just
Feature activation+0.044
Top resid features:
because
Token because
Feature activation+3.742
Top resid features:
new
Token new
Feature activation+0.015
Top resid features:
tools
Token tools
Feature activation-0.212
Top resid features:
arrive
Token arrive
Feature activation+0.202
Top resid features:
,
Token,
Feature activation+0.000
Top resid features:
like
Token like
Feature activation+0.000
Top resid features:
others
Token others
Feature activation-0.048
Top resid features:
found
Token found
Feature activation-0.023
Top resid features:
before
Token before
Feature activation-0.055
Top resid features:
âĢĵ
Token âĢĵ
Feature activation-0.014
Top resid features:
just
Token just
Feature activation-0.015
Top resid features:
because
Token because
Feature activation+3.280
Top resid features:
the
Token the
Feature activation+0.561
Top resid features:
relevant
Token relevant
Feature activation+0.475
Top resid features:
information
Token information
Feature activation-0.182
Top resid features:
is
Token is
Feature activation+0.028
Top resid features:
buried
Token buried
Feature activation+0.000
Top resid features:
others
Token others
Feature activation-0.037
Top resid features:
found
Token found
Feature activation-0.056
Top resid features:
before
Token before
Feature activation-0.146
Top resid features:
âĢĵ
Token âĢĵ
Feature activation-0.014
Top resid features:
just
Token just
Feature activation+0.006
Top resid features:
because
Token because
Feature activation+3.340
Top resid features:
the
Token the
Feature activation+0.724
Top resid features:
relevant
Token relevant
Feature activation+0.636
Top resid features:
information
Token information
Feature activation-0.129
Top resid features:
is
Token is
Feature activation+0.000
Top resid features:
buried
Token buried
Feature activation+0.000
Top resid features:
we
Token we
Feature activation+0.000
Top resid features:
always
Token always
Feature activation-0.050
Top resid features:
see
Token see
Feature activation+0.046
Top resid features:
:
Token:
Feature activation+0.031
Top resid features:
Just
Token Just
Feature activation+0.038
Top resid features:
because
Token because
Feature activation+2.717
Top resid features:
the
Token the
Feature activation+0.805
Top resid features:
defense
Token defense
Feature activation-0.002
Top resid features:
finally
Token finally
Feature activation+0.000
Top resid features:
conceded
Token conceded
Feature activation+0.000
Top resid features:
a
Token a
Feature activation+0.000
Top resid features:
would
Token would
Feature activation-0.029
Top resid features:
genocide
Token genocide
Feature activation-0.062
Top resid features:
become
Token become
Feature activation+0.001
Top resid features:
acceptable
Token acceptable
Feature activation+0.024
Top resid features:
just
Token just
Feature activation+0.011
Top resid features:
because
Token because
Feature activation+3.621
Top resid features:
God
Token God
Feature activation-0.017
Top resid features:
says
Token says
Feature activation+0.025
Top resid features:
so
Token so
Feature activation+0.000
Top resid features:
?
Token?
Feature activation+0.000
Top resid features:
Ċ
TokenĊ
Feature activation+0.000
Top resid features:
oting
Tokenoting
Feature activation-0.006
Top resid features:
for
Token for
Feature activation+0.264
Top resid features:
Roy
Token Roy
Feature activation+0.099
Top resid features:
Moore
Token Moore
Feature activation+0.140
Top resid features:
just
Token just
Feature activation+0.075
Top resid features:
because
Token because
Feature activation+2.555
Top resid features:
he
Token he
Feature activation+0.281
Top resid features:
has
Token has
Feature activation+0.000
Top resid features:
an
Token an
Feature activation+0.000
Top resid features:
âĢ
Token âĢ
Feature activation+0.000
Top resid features:
ĺ
Tokenĺ
Feature activation+0.000
Top resid features:
would
Token would
Feature activation+0.002
Top resid features:
genocide
Token genocide
Feature activation-0.096
Top resid features:
become
Token become
Feature activation+0.040
Top resid features:
acceptable
Token acceptable
Feature activation+0.089
Top resid features:
just
Token just
Feature activation+0.038
Top resid features:
because
Token because
Feature activation+2.751
Top resid features:
God
Token God
Feature activation+0.052
Top resid features:
says
Token says
Feature activation+0.000
Top resid features:
so
Token so
Feature activation+0.000
Top resid features:
?
Token?
Feature activation+0.000
Top resid features:
Ċ
TokenĊ
Feature activation+0.000
Top resid features:
<|endoftext|>
Token<|endoftext|>
Feature activation+0.777
Top resid features:
that
Token that
Feature activation-0.124
Top resid features:
just
Token just
Feature activation+0.092
Top resid features:
because
Token because
Feature activation+1.796
Top resid features:
you
Token you
Feature activation+0.349
Top resid features:
can
Token can
Feature activation+0.842
Top resid features:
do
Token do
Feature activation+0.100
Top resid features:
something
Token something
Feature activation-0.071
Top resid features:
well
Token well
Feature activation+0.000
Top resid features:
It
Token It
Feature activation-0.006
Top resid features:
was
Token was
Feature activation-0.001
Top resid features:
always
Token always
Feature activation-0.054
Top resid features:
just
Token just
Feature activation+0.017
Top resid features:
assumed
Token assumed
Feature activation+0.167
Top resid features:
because
Token because
Feature activation+1.755
Top resid features:
I
Token I
Feature activation+0.581
Top resid features:
was
Token was
Feature activation+0.220
Top resid features:
part
Token part
Feature activation+0.000
Top resid features:
of
Token of
Feature activation+0.000
Top resid features:
a
Token a
Feature activation+0.000
Top resid features:
rule
Token rule
Feature activation-0.084
Top resid features:
just
Token just
Feature activation+0.006
Top resid features:
once
Token once
Feature activation-0.017
Top resid features:
)
Token)
Feature activation+0.035
Top resid features:
That
TokenThat
Feature activation+0.319
Top resid features:
because
Token because
Feature activation+1.861
Top resid features:
I
Token I
Feature activation+0.486
Top resid features:
don
Token don
Feature activation+0.241
Top resid features:
't
Token't
Feature activation+0.063
Top resid features:
consider
Token consider
Feature activation+0.000
Top resid features:
any
Token any
Feature activation+0.000
Top resid features:
any
Token any
Feature activation-0.005
Top resid features:
at
Token at
Feature activation-0.043
Top resid features:
all
Token all
Feature activation-0.123
Top resid features:
,
Token,
Feature activation-0.042
Top resid features:
simply
Token simply
Feature activation+0.114
Top resid features:
because
Token because
Feature activation+2.785
Top resid features:
I
Token I
Feature activation+0.207
Top resid features:
had
Token had
Feature activation+0.000
Top resid features:
other
Token other
Feature activation+0.000
Top resid features:
books
Token books
Feature activation+0.000
Top resid features:
at
Token at
Feature activation+0.000
Top resid features:
others
Token others
Feature activation-0.171
Top resid features:
found
Token found
Feature activation-0.048
Top resid features:
before
Token before
Feature activation-0.125
Top resid features:
âĢĵ
Token âĢĵ
Feature activation-0.117
Top resid features:
just
Token just
Feature activation+0.164
Top resid features:
because
Token because
Feature activation+3.017
Top resid features:
the
Token the
Feature activation+0.179
Top resid features:
relevant
Token relevant
Feature activation+0.000
Top resid features:
information
Token information
Feature activation+0.000
Top resid features:
is
Token is
Feature activation+0.000
Top resid features:
buried
Token buried
Feature activation+0.000
Top resid features:
rule
Token rule
Feature activation-0.121
Top resid features:
just
Token just
Feature activation+0.013
Top resid features:
once
Token once
Feature activation-0.027
Top resid features:
)
Token)
Feature activation+0.048
Top resid features:
That
TokenThat
Feature activation+0.453
Top resid features:
because
Token because
Feature activation+2.154
Top resid features:
I
Token I
Feature activation+0.140
Top resid features:
don
Token don
Feature activation+0.000
Top resid features:
't
Token't
Feature activation+0.000
Top resid features:
consider
Token consider
Feature activation+0.000
Top resid features:
any
Token any
Feature activation+0.000
Top resid features:

Decoder Weights Distribution

Head 0: 0.06

Head 1: 0.05

Head 2: 0.02

Head 3: 0.09

Head 4: 0.06

Head 5: 0.06

Head 6: 0.04

Head 7: 0.02

Head 8: 0.05

Head 9: 0.52

Head 10: 0.02

Head 11: 0.01

Positive logits

teness2.04

past1.92

エル1.78

initially1.76

exists1.75

Range1.74

mentions1.73

receives1.73

ISBN1.66

Slash1.63

ドラ1.63

Schne1.63

participated1.63

Numbers1.62

thouse1.61

existed1.61

zes1.61

lasts1.59

Range1.59

liked1.59

Negative logits

onsense-2.17

srfAttach-2.15

illi-2.09

reconc-1.98

bernatorial-1.89

anners-1.84

-1.83

regretted-1.82

kefeller-1.80

princ-1.76

GBT-1.71

duty-1.70

akia-1.68

Instead-1.68

Interested-1.68

Petty-1.67

oya-1.66

hypocrisy-1.66

SPONSORED-1.65

subsequ-1.64

INTERVAL 4.672 - 5.191
CONTAINS 0.000%

Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
But
TokenBut
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
hammer
Token hammer
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000

INTERVAL 4.153 - 4.672
CONTAINS 0.000%

INTERVAL 3.634 - 4.153
CONTAINS 0.000%

because
Token because
Feature activation+0.000
you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
hammer
Token hammer
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000

INTERVAL 3.115 - 3.634
CONTAINS 0.000%

<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
that
Token that
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.598
you
Token you
Feature activation+5.191
can
Token can
Feature activation+3.451
do
Token do
Feature activation+3.222
something
Token something
Feature activation+1.058
well
Token well
Feature activation+0.000
,
Token,
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
that
Token that
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.598
you
Token you
Feature activation+5.191
can
Token can
Feature activation+3.451
do
Token do
Feature activation+3.222
something
Token something
Feature activation+1.058
well
Token well
Feature activation+0.000
,
Token,
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000

INTERVAL 2.596 - 3.115
CONTAINS 0.000%

âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
new
Token new
Feature activation+0.575
tools
Token tools
Feature activation+2.712
arrive
Token arrive
Feature activation+1.842
,
Token,
Feature activation+0.257
like
Token like
Feature activation+0.000
3
Token 3
Feature activation+0.000
-
Token-
Feature activation+0.000

INTERVAL 2.077 - 2.596
CONTAINS 0.000%

you
Token you
Feature activation+5.049
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.167
ve
Tokenve
Feature activation+4.686
got
Token got
Feature activation+3.800
a
Token a
Feature activation+2.417
hammer
Token hammer
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
t
Tokent
Feature activation+0.000

INTERVAL 1.557 - 2.077
CONTAINS 0.001%

Ŀ
TokenĿ
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
new
Token new
Feature activation+0.575
tools
Token tools
Feature activation+2.712
arrive
Token arrive
Feature activation+1.842
,
Token,
Feature activation+0.257
like
Token like
Feature activation+0.000
3
Token 3
Feature activation+0.000
-
Token-
Feature activation+0.000
D
TokenD
Feature activation+0.000
see
Token see
Feature activation+0.000
:
Token:
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.538
defense
Token defense
Feature activation+1.681
finally
Token finally
Feature activation+0.680
conceded
Token conceded
Feature activation+0.427
a
Token a
Feature activation+0.000
goal
Token goal
Feature activation+0.000
is
Token is
Feature activation+0.000
âĢĵ
Token âĢĵ
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.794
relevant
Token relevant
Feature activation+0.000
information
Token information
Feature activation+1.703
is
Token is
Feature activation+1.707
buried
Token buried
Feature activation+0.000
in
Token in
Feature activation+0.000
a
Token a
Feature activation+0.000
giant
Token giant
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.794
relevant
Token relevant
Feature activation+0.000
information
Token information
Feature activation+1.703
is
Token is
Feature activation+1.707
buried
Token buried
Feature activation+0.000
in
Token in
Feature activation+0.000
a
Token a
Feature activation+0.000
giant
Token giant
Feature activation+0.000
pile
Token pile
Feature activation+0.000
become
Token become
Feature activation+0.000
acceptable
Token acceptable
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
God
Token God
Feature activation+1.061
says
Token says
Feature activation+1.641
so
Token so
Feature activation+0.000
?
Token?
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
If
TokenIf
Feature activation+0.000

INTERVAL 1.038 - 1.557
CONTAINS 0.000%

just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.598
you
Token you
Feature activation+5.191
can
Token can
Feature activation+3.451
do
Token do
Feature activation+3.222
something
Token something
Feature activation+1.058
well
Token well
Feature activation+0.000
,
Token,
Feature activation+0.000
doesn
Token doesn
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
for
Token for
Feature activation+0.000
Roy
Token Roy
Feature activation+0.000
Moore
Token Moore
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
he
Token he
Feature activation+1.144
has
Token has
Feature activation+0.148
an
Token an
Feature activation+0.000
âĢ
Token âĢ
Feature activation+0.000
ĺ
Tokenĺ
Feature activation+0.000
R
TokenR
Feature activation+0.000
genocide
Token genocide
Feature activation+0.000
become
Token become
Feature activation+0.000
acceptable
Token acceptable
Feature activation+0.000
just
Token just
Feature activation+0.000
because
Token because
Feature activation+0.000
God
Token God
Feature activation+1.061
says
Token says
Feature activation+1.641
so
Token so
Feature activation+0.000
?
Token?
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000

INTERVAL 0.519 - 1.038
CONTAINS 0.001%

the
Token the
Feature activation+0.000
fact
Token fact
Feature activation+0.000
that
Token that
Feature activation+0.000
the
Token the
Feature activation+0.000
Koch
Token Koch
Feature activation+0.000
brothers
Token brothers
Feature activation+0.624
are
Token are
Feature activation+0.616
relatively
Token relatively
Feature activation+0.000
unknown
Token unknown
Feature activation+0.000
that
Token that
Feature activation+0.000
allows
Token allows
Feature activation+0.000
always
Token always
Feature activation+0.000
see
Token see
Feature activation+0.000
:
Token:
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.538
defense
Token defense
Feature activation+1.681
finally
Token finally
Feature activation+0.680
conceded
Token conceded
Feature activation+0.427
a
Token a
Feature activation+0.000
goal
Token goal
Feature activation+0.000
.
Token.
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
new
Token new
Feature activation+0.575
tools
Token tools
Feature activation+2.712
arrive
Token arrive
Feature activation+1.842
,
Token,
Feature activation+0.257
like
Token like
Feature activation+0.000
3
Token 3
Feature activation+0.000
fact
Token fact
Feature activation+0.000
that
Token that
Feature activation+0.000
the
Token the
Feature activation+0.000
Koch
Token Koch
Feature activation+0.000
brothers
Token brothers
Feature activation+0.624
are
Token are
Feature activation+0.616
relatively
Token relatively
Feature activation+0.000
unknown
Token unknown
Feature activation+0.000
that
Token that
Feature activation+0.000
allows
Token allows
Feature activation+0.000
such
Token such
Feature activation+0.000
:
Token:
Feature activation+0.000
Just
Token Just
Feature activation+0.000
because
Token because
Feature activation+0.000
the
Token the
Feature activation+0.538
defense
Token defense
Feature activation+1.681
finally
Token finally
Feature activation+0.680
conceded
Token conceded
Feature activation+0.427
a
Token a
Feature activation+0.000
goal
Token goal
Feature activation+0.000
is
Token is
Feature activation+0.000
not
Token not
Feature activation+0.000

INTERVAL 0.000 - 0.519
CONTAINS 99.997%

the
Token the
Feature activation+0.000
property
Token property
Feature activation+0.000
when
Token when
Feature activation+0.000
it
Token it
Feature activation+0.000
was
Token was
Feature activation+0.000
classified
Token classified
Feature activation+0.000
as
Token as
Feature activation+0.000
industrial
Token industrial
Feature activation+0.000
land
Token land
Feature activation+0.000
in
Token in
Feature activation+0.000
2011
Token 2011
Feature activation+0.000
About
Token About
Feature activation+0.000
Actual
Token Actual
Feature activation+0.000
Impact
Token Impact
Feature activation+0.000
Of
Token Of
Feature activation+0.000
Wikileaks
Token Wikileaks
Feature activation+0.000
To
Token To
Feature activation+0.000
Bol
Token Bol
Feature activation+0.000
ster
Tokenster
Feature activation+0.000
Legal
Token Legal
Feature activation+0.000
Eff
Token Eff
Feature activation+0.000
ort
Tokenort
Feature activation+0.000
pay
Token pay
Feature activation+0.000
for
Token for
Feature activation+0.000
their
Token their
Feature activation+0.000
purchases
Token purchases
Feature activation+0.000
with
Token with
Feature activation+0.000
Bitcoin
Token Bitcoin
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Bitcoin
TokenBitcoin
Feature activation+0.000
is
Token is
Feature activation+0.000
put
Token put
Feature activation+0.000
all
Token all
Feature activation+0.000
the
Token the
Feature activation+0.000
resources
Token resources
Feature activation+0.000
in
Token in
Feature activation+0.000
one
Token one
Feature activation+0.000
place
Token place
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Blog
TokenBlog
Feature activation+0.000
Racing
Token Racing
Feature activation+0.000
for
Token for
Feature activation+0.000
remembering
Token remembering
Feature activation+0.000
one
Token one
Feature activation+0.000
of
Token of
Feature activation+0.000
our
Token our
Feature activation+0.000
sport
Token sport
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
s
Tokens
Feature activation+0.000
heroes
Token heroes
Feature activation+0.000

BOTTOM ACTIVATIONS
MIN = 0.000

atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
Development
Token Development
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000

Top feature 3 in H7.9: (feature 31787)

TOP ACTIVATIONS
MAX = 3.228

Bitcoin
Token Bitcoin
Feature activation+0.789
offers
Token offers
Feature activation+2.403
a
Token a
Feature activation+1.447
potentially
Token potentially
Feature activation+1.919
fascinating
Token fascinating
Feature activation+2.008
solution
Token solution
Feature activation+3.228
.
Token.
Feature activation+0.000
In
Token In
Feature activation+0.000
fact
Token fact
Feature activation+0.000
,
Token,
Feature activation+0.000
I
Token I
Feature activation+0.000
and
Token and
Feature activation+0.000
scandals
Token scandals
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
angers
Tokenangers
Feature activation+0.000
and
Token and
Feature activation+0.000
scandals
Token scandals
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Obama
Token Obama
Feature activation+0.000
,
Token,
Feature activation+0.000
about
Token about
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+1.765
writes
Token writes
Feature activation+2.664
,
Token,
Feature activation+0.653
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.014
[
Token[
Feature activation+0.704
i
Tokeni
Feature activation+0.047
which
Token which
Feature activation+0.000
home
Token home
Feature activation+0.000
opathy
Tokenopathy
Feature activation+1.015
had
Token had
Feature activation+2.220
no
Token no
Feature activation+2.252
effect
Token effect
Feature activation+2.505
,
Token,
Feature activation+0.000
he
Token he
Feature activation+0.000
let
Token let
Feature activation+0.000
out
Token out
Feature activation+0.000
a
Token a
Feature activation+0.000
music
Token music
Feature activation+0.000
industry
Token industry
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
Bitcoin
Token Bitcoin
Feature activation+0.789
offers
Token offers
Feature activation+2.403
a
Token a
Feature activation+1.447
potentially
Token potentially
Feature activation+1.919
fascinating
Token fascinating
Feature activation+2.008
solution
Token solution
Feature activation+3.228
.
Token.
Feature activation+0.000
mindless
Token mindless
Feature activation+0.000
ways
Token ways
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
no
Token no
Feature activation+1.570
justification
Token justification
Feature activation+2.355
exists
Token exists
Feature activation+0.819
.
Token.
Feature activation+0.000
Moreover
Token Moreover
Feature activation+0.000
,
Token,
Feature activation+0.000
they
Token they
Feature activation+0.000
alcohol
Token alcohol
Feature activation+0.000
,
Token,
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+2.231
was
Token was
Feature activation+2.353
reimb
Token reimb
Feature activation+0.402
ursed
Tokenursed
Feature activation+1.732
.
Token.
Feature activation+0.000
K
Token K
Feature activation+0.000
ony
Tokenony
Feature activation+0.000
ments
Tokenments
Feature activation+0.000
-
Token -
Feature activation+0.000
on
Token on
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
government
Token government
Feature activation+2.291
will
Token will
Feature activation+2.043
depend
Token depend
Feature activation+1.759
for
Token for
Feature activation+0.253
calculating
Token calculating
Feature activation+0.569
how
Token how
Feature activation+0.000
scandals
Token scandals
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
home
Token home
Feature activation+0.000
opathy
Tokenopathy
Feature activation+1.015
had
Token had
Feature activation+2.220
no
Token no
Feature activation+2.252
effect
Token effect
Feature activation+2.505
,
Token,
Feature activation+0.000
he
Token he
Feature activation+0.000
let
Token let
Feature activation+0.000
out
Token out
Feature activation+0.000
of
Token of
Feature activation+0.000
alcohol
Token alcohol
Feature activation+0.000
,
Token,
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+2.231
was
Token was
Feature activation+2.353
reimb
Token reimb
Feature activation+0.402
ursed
Tokenursed
Feature activation+1.732
.
Token.
Feature activation+0.000
K
Token K
Feature activation+0.000
ailments
Token ailments
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
home
Token home
Feature activation+0.000
opathy
Tokenopathy
Feature activation+1.015
had
Token had
Feature activation+2.220
no
Token no
Feature activation+2.252
effect
Token effect
Feature activation+2.505
,
Token,
Feature activation+0.000
he
Token he
Feature activation+0.000
let
Token let
Feature activation+0.000
sophistication
Token sophistication
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
brand
Token brand
Feature activation+1.600
is
Token is
Feature activation+2.217
known
Token known
Feature activation+0.795
in
Token in
Feature activation+0.000
the
Token the
Feature activation+0.000
hotel
Token hotel
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
women
Token women
Feature activation+0.000
âĢĵ
Token âĢĵ
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+1.939
is
Token is
Feature activation+2.161
yet
Token yet
Feature activation+0.025
to
Token to
Feature activation+1.793
be
Token be
Feature activation+1.705
charged
Token charged
Feature activation+1.666
after
Token after
Feature activation+0.000
procurement
Token procurement
Feature activation+0.000
âĢĶ
Token âĢĶ
Feature activation+0.000
ideas
Token ideas
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+2.065
repeatedly
Token repeatedly
Feature activation+0.000
thanked
Token thanked
Feature activation+0.008
European
Token European
Feature activation+0.000
Commission
Token Commission
Feature activation+0.000
President
Token President
Feature activation+0.000
-
Token -
Feature activation+0.000
on
Token on
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
government
Token government
Feature activation+2.291
will
Token will
Feature activation+2.043
depend
Token depend
Feature activation+1.759
for
Token for
Feature activation+0.253
calculating
Token calculating
Feature activation+0.569
how
Token how
Feature activation+0.000
much
Token much
Feature activation+0.000
which
Token which
Feature activation+0.000
Bitcoin
Token Bitcoin
Feature activation+0.789
offers
Token offers
Feature activation+2.403
a
Token a
Feature activation+1.447
potentially
Token potentially
Feature activation+1.919
fascinating
Token fascinating
Feature activation+2.008
solution
Token solution
Feature activation+3.228
.
Token.
Feature activation+0.000
In
Token In
Feature activation+0.000
fact
Token fact
Feature activation+0.000
,
Token,
Feature activation+0.000
could
Token could
Feature activation+1.940
have
Token have
Feature activation+1.312
had
Token had
Feature activation+0.971
a
Token a
Feature activation+0.402
leading
Token leading
Feature activation+0.276
role
Token role
Feature activation+1.968
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
S
TokenS
Feature activation+0.000
amples
Tokenamples
Feature activation+0.000

Top DFA by src position
MAX = 4.930

in
Token in
Feature activation-0.002
Top resid features:
the
Token the
Feature activation+0.001
Top resid features:
music
Token music
Feature activation-0.006
Top resid features:
industry
Token industry
Feature activation+0.004
Top resid features:
for
Token for
Feature activation+0.009
Top resid features:
which
Token which
Feature activation+4.930
Top resid features:
Bitcoin
Token Bitcoin
Feature activation-0.091
Top resid features:
offers
Token offers
Feature activation-0.087
Top resid features:
a
Token a
Feature activation-0.020
Top resid features:
potentially
Token potentially
Feature activation-0.026
Top resid features:
fascinating
Token fascinating
Feature activation-0.107
Top resid features:
h
Tokenh
Feature activation-0.004
Top resid features:
angers
Tokenangers
Feature activation+0.032
Top resid features:
and
Token and
Feature activation+0.017
Top resid features:
scandals
Token scandals
Feature activation+0.095
Top resid features:
for
Token for
Feature activation+0.057
Top resid features:
which
Token which
Feature activation+3.541
Top resid features:
it
Token it
Feature activation+0.119
Top resid features:
would
Token would
Feature activation+0.067
Top resid features:
later
Token later
Feature activation+0.000
Top resid features:
become
Token become
Feature activation+0.000
Top resid features:
famous
Token famous
Feature activation+0.000
Top resid features:
h
Tokenh
Feature activation-0.004
Top resid features:
angers
Tokenangers
Feature activation+0.014
Top resid features:
and
Token and
Feature activation+0.017
Top resid features:
scandals
Token scandals
Feature activation+0.072
Top resid features:
for
Token for
Feature activation+0.049
Top resid features:
which
Token which
Feature activation+3.666
Top resid features:
it
Token it
Feature activation+0.027
Top resid features:
would
Token would
Feature activation+0.000
Top resid features:
later
Token later
Feature activation+0.000
Top resid features:
become
Token become
Feature activation+0.000
Top resid features:
famous
Token famous
Feature activation+0.000
Top resid features:
by
Token by
Feature activation+0.025
Top resid features:
President
Token President
Feature activation-0.001
Top resid features:
Obama
Token Obama
Feature activation-0.012
Top resid features:
,
Token,
Feature activation+0.034
Top resid features:
about
Token about
Feature activation+0.001
Top resid features:
which
Token which
Feature activation+3.319
Top resid features:
he
Token he
Feature activation+0.085
Top resid features:
writes
Token writes
Feature activation+0.073
Top resid features:
,
Token,
Feature activation+0.000
Top resid features:
âĢ
Token âĢ
Feature activation+0.000
Top resid features:
ľ
Tokenľ
Feature activation+0.000
Top resid features:
a
Token a
Feature activation+0.011
Top resid features:
list
Token list
Feature activation+0.003
Top resid features:
of
Token of
Feature activation-0.014
Top resid features:
ailments
Token ailments
Feature activation+0.078
Top resid features:
for
Token for
Feature activation+0.058
Top resid features:
which
Token which
Feature activation+3.891
Top resid features:
home
Token home
Feature activation-0.023
Top resid features:
opathy
Tokenopathy
Feature activation-0.061
Top resid features:
had
Token had
Feature activation+0.189
Top resid features:
no
Token no
Feature activation-0.029
Top resid features:
effect
Token effect
Feature activation-0.069
Top resid features:
in
Token in
Feature activation-0.003
Top resid features:
the
Token the
Feature activation-0.028
Top resid features:
music
Token music
Feature activation-0.055
Top resid features:
industry
Token industry
Feature activation+0.007
Top resid features:
for
Token for
Feature activation+0.026
Top resid features:
which
Token which
Feature activation+3.687
Top resid features:
Bitcoin
Token Bitcoin
Feature activation-0.071
Top resid features:
offers
Token offers
Feature activation-0.018
Top resid features:
a
Token a
Feature activation+0.000
Top resid features:
potentially
Token potentially
Feature activation+0.000
Top resid features:
fascinating
Token fascinating
Feature activation+0.000
Top resid features:
in
Token in
Feature activation+0.014
Top resid features:
brutally
Token brutally
Feature activation-0.003
Top resid features:
mindless
Token mindless
Feature activation-0.006
Top resid features:
ways
Token ways
Feature activation-0.006
Top resid features:
for
Token for
Feature activation-0.021
Top resid features:
which
Token which
Feature activation+3.362
Top resid features:
no
Token no
Feature activation+0.228
Top resid features:
justification
Token justification
Feature activation-0.144
Top resid features:
exists
Token exists
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
Moreover
Token Moreover
Feature activation+0.000
Top resid features:
worth
Token worth
Feature activation+0.022
Top resid features:
of
Token of
Feature activation+0.002
Top resid features:
alcohol
Token alcohol
Feature activation+0.016
Top resid features:
,
Token,
Feature activation+0.031
Top resid features:
for
Token for
Feature activation+0.031
Top resid features:
which
Token which
Feature activation+3.115
Top resid features:
he
Token he
Feature activation+0.120
Top resid features:
was
Token was
Feature activation+0.072
Top resid features:
reimb
Token reimb
Feature activation+0.000
Top resid features:
ursed
Tokenursed
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
ass
Tokenass
Feature activation+0.001
Top resid features:
ess
Tokeness
Feature activation-0.005
Top resid features:
ments
Tokenments
Feature activation+0.019
Top resid features:
-
Token -
Feature activation-0.022
Top resid features:
on
Token on
Feature activation+0.026
Top resid features:
which
Token which
Feature activation+3.008
Top resid features:
the
Token the
Feature activation+0.150
Top resid features:
government
Token government
Feature activation+0.039
Top resid features:
will
Token will
Feature activation+0.000
Top resid features:
depend
Token depend
Feature activation+0.000
Top resid features:
for
Token for
Feature activation+0.000
Top resid features:
h
Tokenh
Feature activation-0.003
Top resid features:
angers
Tokenangers
Feature activation+0.025
Top resid features:
and
Token and
Feature activation+0.010
Top resid features:
scandals
Token scandals
Feature activation+0.106
Top resid features:
for
Token for
Feature activation+0.032
Top resid features:
which
Token which
Feature activation+2.921
Top resid features:
it
Token it
Feature activation+0.076
Top resid features:
would
Token would
Feature activation+0.146
Top resid features:
later
Token later
Feature activation+0.033
Top resid features:
become
Token become
Feature activation+0.000
Top resid features:
famous
Token famous
Feature activation+0.000
Top resid features:
h
Tokenh
Feature activation-0.005
Top resid features:
angers
Tokenangers
Feature activation+0.028
Top resid features:
and
Token and
Feature activation+0.011
Top resid features:
scandals
Token scandals
Feature activation+0.059
Top resid features:
for
Token for
Feature activation+0.016
Top resid features:
which
Token which
Feature activation+2.822
Top resid features:
it
Token it
Feature activation+0.098
Top resid features:
would
Token would
Feature activation+0.043
Top resid features:
later
Token later
Feature activation+0.204
Top resid features:
become
Token become
Feature activation+0.037
Top resid features:
famous
Token famous
Feature activation+0.000
Top resid features:
a
Token a
Feature activation+0.018
Top resid features:
list
Token list
Feature activation-0.003
Top resid features:
of
Token of
Feature activation+0.064
Top resid features:
ailments
Token ailments
Feature activation+0.079
Top resid features:
for
Token for
Feature activation+0.074
Top resid features:
which
Token which
Feature activation+3.077
Top resid features:
home
Token home
Feature activation-0.039
Top resid features:
opathy
Tokenopathy
Feature activation-0.159
Top resid features:
had
Token had
Feature activation+0.277
Top resid features:
no
Token no
Feature activation+0.076
Top resid features:
effect
Token effect
Feature activation+0.000
Top resid features:
worth
Token worth
Feature activation+0.021
Top resid features:
of
Token of
Feature activation+0.008
Top resid features:
alcohol
Token alcohol
Feature activation-0.019
Top resid features:
,
Token,
Feature activation+0.046
Top resid features:
for
Token for
Feature activation+0.039
Top resid features:
which
Token which
Feature activation+3.158
Top resid features:
he
Token he
Feature activation+0.055
Top resid features:
was
Token was
Feature activation+0.000
Top resid features:
reimb
Token reimb
Feature activation+0.000
Top resid features:
ursed
Tokenursed
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
a
Token a
Feature activation+0.016
Top resid features:
list
Token list
Feature activation-0.039
Top resid features:
of
Token of
Feature activation+0.074
Top resid features:
ailments
Token ailments
Feature activation+0.065
Top resid features:
for
Token for
Feature activation+0.101
Top resid features:
which
Token which
Feature activation+3.152
Top resid features:
home
Token home
Feature activation-0.015
Top resid features:
opathy
Tokenopathy
Feature activation-0.082
Top resid features:
had
Token had
Feature activation+0.095
Top resid features:
no
Token no
Feature activation+0.000
Top resid features:
effect
Token effect
Feature activation+0.000
Top resid features:
,
Token,
Feature activation-0.003
Top resid features:
comfort
Token comfort
Feature activation-0.010
Top resid features:
and
Token and
Feature activation-0.008
Top resid features:
sophistication
Token sophistication
Feature activation-0.020
Top resid features:
for
Token for
Feature activation-0.020
Top resid features:
which
Token which
Feature activation+3.433
Top resid features:
the
Token the
Feature activation+0.097
Top resid features:
brand
Token brand
Feature activation-0.111
Top resid features:
is
Token is
Feature activation+0.025
Top resid features:
known
Token known
Feature activation+0.000
Top resid features:
in
Token in
Feature activation+0.000
Top resid features:
involving
Token involving
Feature activation+0.046
Top resid features:
two
Token two
Feature activation+0.025
Top resid features:
women
Token women
Feature activation-0.006
Top resid features:
âĢĵ
Token âĢĵ
Feature activation-0.022
Top resid features:
for
Token for
Feature activation+0.025
Top resid features:
which
Token which
Feature activation+2.943
Top resid features:
he
Token he
Feature activation+0.099
Top resid features:
is
Token is
Feature activation+0.054
Top resid features:
yet
Token yet
Feature activation+0.000
Top resid features:
to
Token to
Feature activation+0.000
Top resid features:
be
Token be
Feature activation+0.000
Top resid features:
joint
Token joint
Feature activation-0.001
Top resid features:
procurement
Token procurement
Feature activation+0.015
Top resid features:
âĢĶ
Token âĢĶ
Feature activation-0.017
Top resid features:
ideas
Token ideas
Feature activation+0.045
Top resid features:
for
Token for
Feature activation+0.068
Top resid features:
which
Token which
Feature activation+3.067
Top resid features:
he
Token he
Feature activation+0.033
Top resid features:
repeatedly
Token repeatedly
Feature activation+0.000
Top resid features:
thanked
Token thanked
Feature activation+0.000
Top resid features:
European
Token European
Feature activation+0.000
Top resid features:
Commission
Token Commission
Feature activation+0.000
Top resid features:
ass
Tokenass
Feature activation+0.001
Top resid features:
ess
Tokeness
Feature activation-0.009
Top resid features:
ments
Tokenments
Feature activation+0.008
Top resid features:
-
Token -
Feature activation-0.050
Top resid features:
on
Token on
Feature activation+0.034
Top resid features:
which
Token which
Feature activation+2.697
Top resid features:
the
Token the
Feature activation+0.143
Top resid features:
government
Token government
Feature activation+0.062
Top resid features:
will
Token will
Feature activation+0.035
Top resid features:
depend
Token depend
Feature activation+0.000
Top resid features:
for
Token for
Feature activation+0.000
Top resid features:
in
Token in
Feature activation+0.010
Top resid features:
the
Token the
Feature activation-0.001
Top resid features:
music
Token music
Feature activation-0.027
Top resid features:
industry
Token industry
Feature activation+0.014
Top resid features:
for
Token for
Feature activation+0.033
Top resid features:
which
Token which
Feature activation+3.287
Top resid features:
Bitcoin
Token Bitcoin
Feature activation-0.064
Top resid features:
offers
Token offers
Feature activation-0.007
Top resid features:
a
Token a
Feature activation+0.020
Top resid features:
potentially
Token potentially
Feature activation-0.012
Top resid features:
fascinating
Token fascinating
Feature activation+0.006
Top resid features:
North
Token North
Feature activation-0.011
Top resid features:
Fr
Token Fr
Feature activation-0.004
Top resid features:
is
Tokenis
Feature activation-0.009
Top resid features:
ia
Tokenia
Feature activation-0.007
Top resid features:
for
Token for
Feature activation-0.079
Top resid features:
which
Token which
Feature activation+2.868
Top resid features:
the
Token the
Feature activation+0.056
Top resid features:
dialect
Token dialect
Feature activation-0.020
Top resid features:
could
Token could
Feature activation+0.133
Top resid features:
have
Token have
Feature activation+0.075
Top resid features:
had
Token had
Feature activation+0.154
Top resid features:

Decoder Weights Distribution

Head 0: 0.05

Head 1: 0.03

Head 2: 0.01

Head 3: 0.08

Head 4: 0.04

Head 5: 0.06

Head 6: 0.04

Head 7: 0.02

Head 8: 0.14

Head 9: 0.50

Head 10: 0.01

Head 11: 0.02

Positive logits

��2.13

blush2.11

luaj2.07

40901.87

indifference1.82

reply1.77

Cosponsors1.77

ㅋㅋ1.77

succumbed1.72

fortunately1.71

reply1.71

eloqu1.68

theoret1.68

ooked1.68

sadly1.67

arthed1.66

Benef1.65

cooldown1.65

laim1.65

microsoft1.64

Negative logits

uto-2.03

aneers-1.94

nian-1.89

ividual-1.85

ilda-1.85

unda-1.78

upiter-1.77

-1.76

-1.72

ilet-1.72

otta-1.72

populate-1.71

ulent-1.71

THESE-1.69

SPACE-1.68

deen-1.67

anish-1.67

LOC-1.67

Solitaire-1.66

plaza-1.65

INTERVAL 2.906 - 3.228
CONTAINS 0.000%

Bitcoin
Token Bitcoin
Feature activation+0.789
offers
Token offers
Feature activation+2.403
a
Token a
Feature activation+1.447
potentially
Token potentially
Feature activation+1.919
fascinating
Token fascinating
Feature activation+2.008
solution
Token solution
Feature activation+3.228
.
Token.
Feature activation+0.000
In
Token In
Feature activation+0.000
fact
Token fact
Feature activation+0.000
,
Token,
Feature activation+0.000
I
Token I
Feature activation+0.000

INTERVAL 2.583 - 2.906
CONTAINS 0.000%

angers
Tokenangers
Feature activation+0.000
and
Token and
Feature activation+0.000
scandals
Token scandals
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
and
Token and
Feature activation+0.000
scandals
Token scandals
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Obama
Token Obama
Feature activation+0.000
,
Token,
Feature activation+0.000
about
Token about
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+1.765
writes
Token writes
Feature activation+2.664
,
Token,
Feature activation+0.653
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.014
[
Token[
Feature activation+0.704
i
Tokeni
Feature activation+0.047

INTERVAL 2.260 - 2.583
CONTAINS 0.001%

music
Token music
Feature activation+0.000
industry
Token industry
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
Bitcoin
Token Bitcoin
Feature activation+0.789
offers
Token offers
Feature activation+2.403
a
Token a
Feature activation+1.447
potentially
Token potentially
Feature activation+1.919
fascinating
Token fascinating
Feature activation+2.008
solution
Token solution
Feature activation+3.228
.
Token.
Feature activation+0.000
which
Token which
Feature activation+0.000
home
Token home
Feature activation+0.000
opathy
Tokenopathy
Feature activation+1.015
had
Token had
Feature activation+2.220
no
Token no
Feature activation+2.252
effect
Token effect
Feature activation+2.505
,
Token,
Feature activation+0.000
he
Token he
Feature activation+0.000
let
Token let
Feature activation+0.000
out
Token out
Feature activation+0.000
a
Token a
Feature activation+0.000
alcohol
Token alcohol
Feature activation+0.000
,
Token,
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+2.231
was
Token was
Feature activation+2.353
reimb
Token reimb
Feature activation+0.402
ursed
Tokenursed
Feature activation+1.732
.
Token.
Feature activation+0.000
K
Token K
Feature activation+0.000
ony
Tokenony
Feature activation+0.000
mindless
Token mindless
Feature activation+0.000
ways
Token ways
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
no
Token no
Feature activation+1.570
justification
Token justification
Feature activation+2.355
exists
Token exists
Feature activation+0.819
.
Token.
Feature activation+0.000
Moreover
Token Moreover
Feature activation+0.000
,
Token,
Feature activation+0.000
they
Token they
Feature activation+0.000
scandals
Token scandals
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000

INTERVAL 1.937 - 2.260
CONTAINS 0.001%

of
Token of
Feature activation+0.000
alcohol
Token alcohol
Feature activation+0.000
,
Token,
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+2.231
was
Token was
Feature activation+2.353
reimb
Token reimb
Feature activation+0.402
ursed
Tokenursed
Feature activation+1.732
.
Token.
Feature activation+0.000
K
Token K
Feature activation+0.000
-
Token -
Feature activation+0.000
on
Token on
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
government
Token government
Feature activation+2.291
will
Token will
Feature activation+2.043
depend
Token depend
Feature activation+1.759
for
Token for
Feature activation+0.253
calculating
Token calculating
Feature activation+0.569
how
Token how
Feature activation+0.000
much
Token much
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
sophistication
Token sophistication
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
brand
Token brand
Feature activation+1.600
is
Token is
Feature activation+2.217
known
Token known
Feature activation+0.795
in
Token in
Feature activation+0.000
the
Token the
Feature activation+0.000
hotel
Token hotel
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
procurement
Token procurement
Feature activation+0.000
âĢĶ
Token âĢĶ
Feature activation+0.000
ideas
Token ideas
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+2.065
repeatedly
Token repeatedly
Feature activation+0.000
thanked
Token thanked
Feature activation+0.008
European
Token European
Feature activation+0.000
Commission
Token Commission
Feature activation+0.000
President
Token President
Feature activation+0.000

INTERVAL 1.614 - 1.937
CONTAINS 0.001%

President
Token President
Feature activation+0.000
Obama
Token Obama
Feature activation+0.000
,
Token,
Feature activation+0.000
about
Token about
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+1.765
writes
Token writes
Feature activation+2.664
,
Token,
Feature activation+0.653
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.014
[
Token[
Feature activation+0.704
on
Token on
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
government
Token government
Feature activation+2.291
will
Token will
Feature activation+2.043
depend
Token depend
Feature activation+1.759
for
Token for
Feature activation+0.253
calculating
Token calculating
Feature activation+0.569
how
Token how
Feature activation+0.000
much
Token much
Feature activation+0.000
carbon
Token carbon
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
Bitcoin
Token Bitcoin
Feature activation+0.789
offers
Token offers
Feature activation+2.403
a
Token a
Feature activation+1.447
potentially
Token potentially
Feature activation+1.919
fascinating
Token fascinating
Feature activation+2.008
solution
Token solution
Feature activation+3.228
.
Token.
Feature activation+0.000
In
Token In
Feature activation+0.000
fact
Token fact
Feature activation+0.000
which
Token which
Feature activation+0.000
it
Token it
Feature activation+2.706
would
Token would
Feature activation+2.833
later
Token later
Feature activation+2.284
become
Token become
Feature activation+2.253
famous
Token famous
Feature activation+1.879
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
tone
Token tone
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+1.939
is
Token is
Feature activation+2.161
yet
Token yet
Feature activation+0.025
to
Token to
Feature activation+1.793
be
Token be
Feature activation+1.705
charged
Token charged
Feature activation+1.666
after
Token after
Feature activation+0.000
four
Token four
Feature activation+0.000
years
Token years
Feature activation+0.000
.
Token.
Feature activation+0.000

INTERVAL 1.291 - 1.614
CONTAINS 0.001%

for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.637
dialect
Token dialect
Feature activation+0.683
could
Token could
Feature activation+1.940
have
Token have
Feature activation+1.312
had
Token had
Feature activation+0.971
a
Token a
Feature activation+0.402
leading
Token leading
Feature activation+0.276
role
Token role
Feature activation+1.968
.
Token.
Feature activation+0.000
of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
we
Token we
Feature activation+1.004
are
Token are
Feature activation+1.039
all
Token all
Feature activation+1.285
so
Token so
Feature activation+1.307
very
Token very
Feature activation+1.274
,
Token,
Feature activation+0.000
very
Token very
Feature activation+0.797
proud
Token proud
Feature activation+1.096
.
Token.
Feature activation+0.000
brutally
Token brutally
Feature activation+0.000
mindless
Token mindless
Feature activation+0.000
ways
Token ways
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
no
Token no
Feature activation+1.570
justification
Token justification
Feature activation+2.355
exists
Token exists
Feature activation+0.819
.
Token.
Feature activation+0.000
Moreover
Token Moreover
Feature activation+0.000
,
Token,
Feature activation+0.000
and
Token and
Feature activation+0.000
sophistication
Token sophistication
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
brand
Token brand
Feature activation+1.600
is
Token is
Feature activation+2.217
known
Token known
Feature activation+0.795
in
Token in
Feature activation+0.000
the
Token the
Feature activation+0.000
hotel
Token hotel
Feature activation+0.000
with
Token with
Feature activation+0.000
which
Token which
Feature activation+0.000
I
Token I
Feature activation+0.776
tend
Token tend
Feature activation+0.716
to
Token to
Feature activation+0.729
agree
Token agree
Feature activation+1.544
:
Token:
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Quote
TokenQuote
Feature activation+0.000
W
Token W
Feature activation+0.000

INTERVAL 0.969 - 1.291
CONTAINS 0.002%

of
Token of
Feature activation+0.000
which
Token which
Feature activation+0.000
nut
Token nut
Feature activation+0.000
items
Token items
Feature activation+0.000
are
Token are
Feature activation+0.726
a
Token a
Feature activation+1.200
viable
Token viable
Feature activation+0.985
source
Token source
Feature activation+1.452
.
Token.
Feature activation+0.000
There
Token There
Feature activation+0.000
are
Token are
Feature activation+0.000
and
Token and
Feature activation+0.000
fantasy
Token fantasy
Feature activation+0.000
characters
Token characters
Feature activation+0.000
with
Token with
Feature activation+0.000
which
Token which
Feature activation+0.000
they
Token they
Feature activation+1.026
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.485
ident
Tokenident
Feature activation+0.000
ify
Tokenify
Feature activation+0.000
,
Token,
Feature activation+0.000
of
Token of
Feature activation+0.000
ailments
Token ailments
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
home
Token home
Feature activation+0.000
opathy
Tokenopathy
Feature activation+1.015
had
Token had
Feature activation+2.220
no
Token no
Feature activation+2.252
effect
Token effect
Feature activation+2.505
,
Token,
Feature activation+0.000
he
Token he
Feature activation+0.000
area
Token area
Feature activation+0.000
had
Token had
Feature activation+0.976
had
Token had
Feature activation+1.053
the
Token the
Feature activation+0.524
most
Token most
Feature activation+0.916
experience
Token experience
Feature activation+1.202
,"
Token,"
Feature activation+0.000
he
Token he
Feature activation+0.000
said
Token said
Feature activation+0.000
.
Token.
Feature activation+0.000
"
Token "
Feature activation+0.000
many
Token many
Feature activation+0.000
different
Token different
Feature activation+0.000
teams
Token teams
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
they
Token they
Feature activation+0.982
played
Token played
Feature activation+0.662
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Green
TokenGreen
Feature activation+0.000

INTERVAL 0.646 - 0.969
CONTAINS 0.003%

are
Token are
Feature activation+1.039
all
Token all
Feature activation+1.285
so
Token so
Feature activation+1.307
very
Token very
Feature activation+1.274
,
Token,
Feature activation+0.000
very
Token very
Feature activation+0.797
proud
Token proud
Feature activation+1.096
.
Token.
Feature activation+0.000
Thank
Token Thank
Feature activation+0.000
you
Token you
Feature activation+0.000
so
Token so
Feature activation+0.000
he
Token he
Feature activation+1.765
writes
Token writes
Feature activation+2.664
,
Token,
Feature activation+0.653
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.014
[
Token[
Feature activation+0.704
i
Tokeni
Feature activation+0.047
]
Token]
Feature activation+0.155
n
Tokenn
Feature activation+0.000
quoting
Token quoting
Feature activation+0.000
his
Token his
Feature activation+0.000
<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
capital
Token capital
Feature activation+0.000
,
Token,
Feature activation+0.000
failing
Token failing
Feature activation+0.000
which
Token which
Feature activation+0.000
its
Token its
Feature activation+0.677
plea
Token plea
Feature activation+0.000
to
Token to
Feature activation+0.000
procure
Token procure
Feature activation+0.000
licence
Token licence
Feature activation+0.000
will
Token will
Feature activation+0.066
different
Token different
Feature activation+0.000
teams
Token teams
Feature activation+0.000
for
Token for
Feature activation+0.000
which
Token which
Feature activation+0.000
they
Token they
Feature activation+0.982
played
Token played
Feature activation+0.662
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Green
TokenGreen
Feature activation+0.000
e
Tokene
Feature activation+0.000
,
Token,
Feature activation+0.000
about
Token about
Feature activation+0.000
which
Token which
Feature activation+0.000
he
Token he
Feature activation+1.765
writes
Token writes
Feature activation+2.664
,
Token,
Feature activation+0.653
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.014
[
Token[
Feature activation+0.704
i
Tokeni
Feature activation+0.047
]
Token]
Feature activation+0.155

INTERVAL 0.323 - 0.646
CONTAINS 0.007%

which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
public
Token public
Feature activation+0.000
may
Token may
Feature activation+0.144
not
Token not
Feature activation+0.000
be
Token be
Feature activation+0.350
fully
Token fully
Feature activation+0.423
aware
Token aware
Feature activation+0.000
,
Token,
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
and
Token and
Feature activation+0.000
to
Token to
Feature activation+0.000
which
Token which
Feature activation+0.000
the
Token the
Feature activation+0.000
dog
Token dog
Feature activation+0.000
responds
Token responds
Feature activation+0.563
appropriately
Token appropriately
Feature activation+0.000
.
Token.
Feature activation+0.000
For
Token For
Feature activation+0.000
example
Token example
Feature activation+0.000
my
Token my
Feature activation+0.000
they
Token they
Feature activation+0.000
nor
Token nor
Feature activation+0.000
we
Token we
Feature activation+0.000
could
Token could
Feature activation+0.243
successfully
Token successfully
Feature activation+0.461
control
Token control
Feature activation+0.543
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
system
Token system
Feature activation+0.000
of
Token of
Feature activation+0.000
social
Token social
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
For
TokenFor
Feature activation+0.000
this
Token this
Feature activation+0.000
I
Token I
Feature activation+0.074
was
Token was
Feature activation+0.443
called
Token called
Feature activation+0.118
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.000
a
Tokena
Feature activation+0.000
Tory
Token Tory
Feature activation+0.000
Director
Token Director
Feature activation+0.000
Jeff
Token Jeff
Feature activation+0.000
K
Token K
Feature activation+0.000
ony
Tokenony
Feature activation+0.000
a
Tokena
Feature activation+0.474
bought
Token bought
Feature activation+0.582
$
Token $
Feature activation+0.000
79
Token79
Feature activation+0.000
worth
Token worth
Feature activation+0.000
of
Token of
Feature activation+0.000
alcohol
Token alcohol
Feature activation+0.000

INTERVAL 0.000 - 0.323
CONTAINS 99.984%

fatal
Token fatal
Feature activation+0.000
for
Token for
Feature activation+0.000
Israel
Token Israel
Feature activation+0.000
as
Token as
Feature activation+0.000
a
Token a
Feature activation+0.000
Jewish
Token Jewish
Feature activation+0.000
and
Token and
Feature activation+0.000
democratic
Token democratic
Feature activation+0.000
state
Token state
Feature activation+0.000
.
Token.
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
hardware
Token hardware
Feature activation+0.000
,
Token,
Feature activation+0.000
or
Token or
Feature activation+0.000
remain
Token remain
Feature activation+0.000
insecure
Token insecure
Feature activation+0.000
if
Token if
Feature activation+0.000
a
Token a
Feature activation+0.000
vulnerable
Token vulnerable
Feature activation+0.000
configuration
Token configuration
Feature activation+0.000
is
Token is
Feature activation+0.000
not
Token not
Feature activation+0.000
Mutant
Token Mutant
Feature activation+0.000
Ninja
Token Ninja
Feature activation+0.000
Turtles
Token Turtles
Feature activation+0.000
2
Token 2
Feature activation+0.000
will
Token will
Feature activation+0.000
hit
Token hit
Feature activation+0.000
theaters
Token theaters
Feature activation+0.000
on
Token on
Feature activation+0.000
June
Token June
Feature activation+0.000
3
Token 3
Feature activation+0.000
,
Token,
Feature activation+0.000
?
Token?
Feature activation+0.000
Smith
Token Smith
Feature activation+0.000
,
Token,
Feature activation+0.000
an
Token an
Feature activation+0.000
aff
Token aff
Feature activation+0.000
able
Tokenable
Feature activation+0.000
self
Token self
Feature activation+0.000
-
Token-
Feature activation+0.000
prom
Tokenprom
Feature activation+0.000
oter
Tokenoter
Feature activation+0.000
,
Token,
Feature activation+0.000
securities
Token securities
Feature activation+0.000
laws
Token laws
Feature activation+0.000
,
Token,
Feature activation+0.000
the
Token the
Feature activation+0.000
attorney
Token attorney
Feature activation+0.000
general
Token general
Feature activation+0.000
is
Token is
Feature activation+0.000
also
Token also
Feature activation+0.000
investigating
Token investigating
Feature activation+0.000
a
Token a
Feature activation+0.000
leading
Token leading
Feature activation+0.000

BOTTOM ACTIVATIONS
MIN = 0.000

atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
Development
Token Development
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000

Top feature 4 in H7.9: (feature 46540)

TOP ACTIVATIONS
MAX = 5.952

least
Token least
Feature activation+0.937
I
Token I
Feature activation+4.081
didn
Token didn
Feature activation+3.322
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
t
Tokent
Feature activation+5.952
die
Token die
Feature activation+0.399
or
Token or
Feature activation+0.000
anything
Token anything
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Well
TokenWell
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.451
I
Token I
Feature activation+4.829
drink
Token drink
Feature activation+1.974
beer
Token beer
Feature activation+0.852
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
a
Token a
Feature activation+0.000
Hey
Token Hey
Feature activation+0.000
ladies
Token ladies
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.306
we
Token we
Feature activation+4.821
know
Token know
Feature activation+3.996
what
Token what
Feature activation+1.646
Donald
Token Donald
Feature activation+0.000
Trump
Token Trump
Feature activation+0.000
thinks
Token thinks
Feature activation+0.362
back
Token back
Feature activation+0.000
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.937
I
Token I
Feature activation+4.081
didn
Token didn
Feature activation+3.322
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
t
Tokent
Feature activation+5.952
die
Token die
Feature activation+0.399
ladies
Token ladies
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.306
we
Token we
Feature activation+4.821
know
Token know
Feature activation+3.996
what
Token what
Feature activation+1.646
Donald
Token Donald
Feature activation+0.000
Trump
Token Trump
Feature activation+0.000
thinks
Token thinks
Feature activation+0.362
of
Token of
Feature activation+0.000
in
Token in
Feature activation+0.000
Brazil
Token Brazil
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+3.838
will
Token will
Feature activation+3.301
return
Token return
Feature activation+0.522
home
Token home
Feature activation+0.621
with
Token with
Feature activation+0.037
clean
Token clean
Feature activation+0.000
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.993
she
Token she
Feature activation+3.099
's
Token's
Feature activation+3.692
slightly
Token slightly
Feature activation+1.975
more
Token more
Feature activation+1.462
rel
Token rel
Feature activation+0.000
atable
Tokenatable
Feature activation+0.997
.
Token.
Feature activation+0.000
least
Token least
Feature activation+0.000
the
Token the
Feature activation+1.859
Lo
Token Lo
Feature activation+0.000
ons
Tokenons
Feature activation+0.853
won
Token won
Feature activation+2.606
't
Token't
Feature activation+3.426
be
Token be
Feature activation+2.775
missing
Token missing
Feature activation+1.577
any
Token any
Feature activation+0.794
key
Token key
Feature activation+0.307
pieces
Token pieces
Feature activation+0.895
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.937
I
Token I
Feature activation+4.081
didn
Token didn
Feature activation+3.322
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
t
Tokent
Feature activation+5.952
die
Token die
Feature activation+0.399
or
Token or
Feature activation+0.000
Brazil
Token Brazil
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+3.838
will
Token will
Feature activation+3.301
return
Token return
Feature activation+0.522
home
Token home
Feature activation+0.621
with
Token with
Feature activation+0.037
clean
Token clean
Feature activation+0.000
cons
Token cons
Feature activation+0.000
At
Token At
Feature activation+0.000
least
Token least
Feature activation+0.000
for
Token for
Feature activation+0.523
me
Token me
Feature activation+0.784
,
Token,
Feature activation+1.456
I
Token I
Feature activation+3.272
mean
Token mean
Feature activation+0.599
,
Token,
Feature activation+1.342
shit
Token shit
Feature activation+0.424
,
Token,
Feature activation+0.707
I
Token I
Feature activation+0.574
hands
Token hands
Feature activation+0.000
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+1.320
they
Token they
Feature activation+3.260
could
Token could
Feature activation+2.459
allocate
Token allocate
Feature activation+0.486
some
Token some
Feature activation+0.343
profits
Token profits
Feature activation+0.816
to
Token to
Feature activation+1.088
AI
Token AI
Feature activation+0.000
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.993
she
Token she
Feature activation+3.099
's
Token's
Feature activation+3.692
slightly
Token slightly
Feature activation+1.975
more
Token more
Feature activation+1.462
rel
Token rel
Feature activation+0.000
atable
Tokenatable
Feature activation+0.997
world
Token world
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
the
Token the
Feature activation+0.166
world
Token world
Feature activation+3.062
inside
Token inside
Feature activation+1.155
the
Token the
Feature activation+0.000
home
Token home
Feature activation+0.000
,
Token,
Feature activation+0.000
I
Token I
Feature activation+0.000
?
Token?
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+2.932
won
Token won
Feature activation+1.973
't
Token't
Feature activation+2.789
be
Token be
Feature activation+2.789
exposed
Token exposed
Feature activation+0.667
to
Token to
Feature activation+0.518
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.846
there
Token there
Feature activation+2.243
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
s
Tokens
Feature activation+2.812
been
Token been
Feature activation+1.253
improvement
Token improvement
Feature activation+0.129
,
Token,
Feature activation+0.443
said
Token said
Feature activation+0.000
David
Token David
Feature activation+0.000
the
Token the
Feature activation+0.000
infantry
Token infantry
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+1.288
not
Token not
Feature activation+2.812
immediately
Token immediately
Feature activation+0.533
,
Token,
Feature activation+0.000
since
Token since
Feature activation+0.000
it
Token it
Feature activation+0.000
remains
Token remains
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+2.932
won
Token won
Feature activation+1.973
't
Token't
Feature activation+2.789
be
Token be
Feature activation+2.789
exposed
Token exposed
Feature activation+0.667
to
Token to
Feature activation+0.518
the
Token the
Feature activation+0.000
"
Token "
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+2.932
won
Token won
Feature activation+1.973
't
Token't
Feature activation+2.789
be
Token be
Feature activation+2.789
exposed
Token exposed
Feature activation+0.667
to
Token to
Feature activation+0.518
the
Token the
Feature activation+0.000
"
Token "
Feature activation+0.000
extreme
Tokenextreme
Feature activation+0.000
the
Token the
Feature activation+1.859
Lo
Token Lo
Feature activation+0.000
ons
Tokenons
Feature activation+0.853
won
Token won
Feature activation+2.606
't
Token't
Feature activation+3.426
be
Token be
Feature activation+2.775
missing
Token missing
Feature activation+1.577
any
Token any
Feature activation+0.794
key
Token key
Feature activation+0.307
pieces
Token pieces
Feature activation+0.895
heading
Token heading
Feature activation+0.000

Top DFA by src position
MAX = 5.922

my
Token my
Feature activation+0.009
Top resid features:
back
Token back
Feature activation-0.001
Top resid features:
,
Token,
Feature activation+0.019
Top resid features:
but
Token but
Feature activation+0.107
Top resid features:
at
Token at
Feature activation+0.054
Top resid features:
least
Token least
Feature activation+5.671
Top resid features:
I
Token I
Feature activation+0.756
Top resid features:
didn
Token didn
Feature activation+0.962
Top resid features:
âĢ
TokenâĢ
Feature activation+0.046
Top resid features:
Ļ
TokenĻ
Feature activation-0.050
Top resid features:
t
Tokent
Feature activation+0.184
Top resid features:
<|endoftext|>
Token<|endoftext|>
Feature activation+0.786
Top resid features:
Ċ
TokenĊ
Feature activation-0.005
Top resid features:
Well
TokenWell
Feature activation+0.029
Top resid features:
,
Token,
Feature activation+0.055
Top resid features:
at
Token at
Feature activation+0.070
Top resid features:
least
Token least
Feature activation+5.841
Top resid features:
I
Token I
Feature activation+0.283
Top resid features:
drink
Token drink
Feature activation+0.000
Top resid features:
beer
Token beer
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.000
Top resid features:
As
Token As
Feature activation+0.000
Top resid features:
:
Token:
Feature activation+0.132
Top resid features:
Hey
Token Hey
Feature activation-0.009
Top resid features:
ladies
Token ladies
Feature activation-0.227
Top resid features:
,
Token,
Feature activation+0.084
Top resid features:
at
Token at
Feature activation+0.069
Top resid features:
least
Token least
Feature activation+5.922
Top resid features:
we
Token we
Feature activation+0.410
Top resid features:
know
Token know
Feature activation+0.000
Top resid features:
what
Token what
Feature activation+0.000
Top resid features:
Donald
Token Donald
Feature activation+0.000
Top resid features:
Trump
Token Trump
Feature activation+0.000
Top resid features:
my
Token my
Feature activation+0.019
Top resid features:
back
Token back
Feature activation+0.006
Top resid features:
,
Token,
Feature activation+0.054
Top resid features:
but
Token but
Feature activation+0.161
Top resid features:
at
Token at
Feature activation+0.102
Top resid features:
least
Token least
Feature activation+5.497
Top resid features:
I
Token I
Feature activation+0.379
Top resid features:
didn
Token didn
Feature activation+0.000
Top resid features:
âĢ
TokenâĢ
Feature activation+0.000
Top resid features:
Ļ
TokenĻ
Feature activation+0.000
Top resid features:
t
Tokent
Feature activation+0.000
Top resid features:
:
Token:
Feature activation+0.045
Top resid features:
Hey
Token Hey
Feature activation+0.029
Top resid features:
ladies
Token ladies
Feature activation-0.078
Top resid features:
,
Token,
Feature activation+0.086
Top resid features:
at
Token at
Feature activation+0.076
Top resid features:
least
Token least
Feature activation+3.748
Top resid features:
we
Token we
Feature activation+1.437
Top resid features:
know
Token know
Feature activation+0.282
Top resid features:
what
Token what
Feature activation+0.000
Top resid features:
Donald
Token Donald
Feature activation+0.000
Top resid features:
Trump
Token Trump
Feature activation+0.000
Top resid features:
fans
Token fans
Feature activation-0.067
Top resid features:
in
Token in
Feature activation+0.021
Top resid features:
Brazil
Token Brazil
Feature activation-0.034
Top resid features:
,
Token,
Feature activation+0.009
Top resid features:
at
Token at
Feature activation+0.120
Top resid features:
least
Token least
Feature activation+4.775
Top resid features:
they
Token they
Feature activation+0.434
Top resid features:
will
Token will
Feature activation+0.000
Top resid features:
return
Token return
Feature activation+0.000
Top resid features:
home
Token home
Feature activation+0.000
Top resid features:
with
Token with
Feature activation+0.000
Top resid features:
of
Token of
Feature activation-0.012
Top resid features:
AI
Token AI
Feature activation-0.031
Top resid features:
,
Token,
Feature activation+0.087
Top resid features:
but
Token but
Feature activation+0.092
Top resid features:
at
Token at
Feature activation+0.107
Top resid features:
least
Token least
Feature activation+3.802
Top resid features:
she
Token she
Feature activation+0.841
Top resid features:
's
Token's
Feature activation+0.252
Top resid features:
slightly
Token slightly
Feature activation+0.000
Top resid features:
more
Token more
Feature activation+0.000
Top resid features:
rel
Token rel
Feature activation+0.000
Top resid features:
been
Token been
Feature activation+0.013
Top resid features:
."
Token."
Feature activation+0.032
Top resid features:
Ċ
TokenĊ
Feature activation+0.039
Top resid features:
Ċ
TokenĊ
Feature activation-0.012
Top resid features:
At
TokenAt
Feature activation+0.072
Top resid features:
least
Token least
Feature activation+3.451
Top resid features:
the
Token the
Feature activation+0.419
Top resid features:
Lo
Token Lo
Feature activation-0.014
Top resid features:
ons
Tokenons
Feature activation+0.064
Top resid features:
won
Token won
Feature activation+0.420
Top resid features:
't
Token't
Feature activation+0.456
Top resid features:
my
Token my
Feature activation+0.018
Top resid features:
back
Token back
Feature activation+0.005
Top resid features:
,
Token,
Feature activation+0.038
Top resid features:
but
Token but
Feature activation+0.119
Top resid features:
at
Token at
Feature activation+0.085
Top resid features:
least
Token least
Feature activation+3.344
Top resid features:
I
Token I
Feature activation+0.929
Top resid features:
didn
Token didn
Feature activation+0.451
Top resid features:
âĢ
TokenâĢ
Feature activation+0.000
Top resid features:
Ļ
TokenĻ
Feature activation+0.000
Top resid features:
t
Tokent
Feature activation+0.000
Top resid features:
fans
Token fans
Feature activation-0.086
Top resid features:
in
Token in
Feature activation+0.018
Top resid features:
Brazil
Token Brazil
Feature activation-0.045
Top resid features:
,
Token,
Feature activation+0.048
Top resid features:
at
Token at
Feature activation+0.116
Top resid features:
least
Token least
Feature activation+2.981
Top resid features:
they
Token they
Feature activation+1.311
Top resid features:
will
Token will
Feature activation+0.323
Top resid features:
return
Token return
Feature activation+0.000
Top resid features:
home
Token home
Feature activation+0.000
Top resid features:
with
Token with
Feature activation+0.000
Top resid features:
cath
Token cath
Feature activation+0.011
Top resid features:
art
Tokenart
Feature activation-0.001
Top resid features:
ic
Tokenic
Feature activation-0.010
Top resid features:
.
Token.
Feature activation+0.067
Top resid features:
At
Token At
Feature activation+0.078
Top resid features:
least
Token least
Feature activation+3.633
Top resid features:
for
Token for
Feature activation+0.201
Top resid features:
me
Token me
Feature activation+0.210
Top resid features:
,
Token,
Feature activation+0.240
Top resid features:
I
Token I
Feature activation+0.133
Top resid features:
mean
Token mean
Feature activation+0.000
Top resid features:
dead
Token dead
Feature activation+0.076
Top resid features:
hands
Token hands
Feature activation-0.202
Top resid features:
,
Token,
Feature activation+0.022
Top resid features:
but
Token but
Feature activation+0.181
Top resid features:
at
Token at
Feature activation+0.140
Top resid features:
least
Token least
Feature activation+4.247
Top resid features:
they
Token they
Feature activation+0.356
Top resid features:
could
Token could
Feature activation+0.000
Top resid features:
allocate
Token allocate
Feature activation+0.000
Top resid features:
some
Token some
Feature activation+0.000
Top resid features:
profits
Token profits
Feature activation+0.000
Top resid features:
of
Token of
Feature activation-0.021
Top resid features:
AI
Token AI
Feature activation-0.067
Top resid features:
,
Token,
Feature activation+0.085
Top resid features:
but
Token but
Feature activation+0.079
Top resid features:
at
Token at
Feature activation+0.089
Top resid features:
least
Token least
Feature activation+4.323
Top resid features:
she
Token she
Feature activation+0.306
Top resid features:
's
Token's
Feature activation+0.000
Top resid features:
slightly
Token slightly
Feature activation+0.000
Top resid features:
more
Token more
Feature activation+0.000
Top resid features:
rel
Token rel
Feature activation+0.000
Top resid features:
to
Token to
Feature activation+0.022
Top resid features:
the
Token the
Feature activation+0.024
Top resid features:
world
Token world
Feature activation-0.020
Top resid features:
,
Token,
Feature activation+0.006
Top resid features:
at
Token at
Feature activation+0.104
Top resid features:
least
Token least
Feature activation+3.515
Top resid features:
the
Token the
Feature activation+0.890
Top resid features:
world
Token world
Feature activation+0.058
Top resid features:
inside
Token inside
Feature activation+0.000
Top resid features:
the
Token the
Feature activation+0.000
Top resid features:
home
Token home
Feature activation+0.000
Top resid features:
play
Token play
Feature activation+0.002
Top resid features:
?
Token?
Feature activation+0.011
Top resid features:
Ċ
TokenĊ
Feature activation+0.029
Top resid features:
Ċ
TokenĊ
Feature activation+0.028
Top resid features:
At
TokenAt
Feature activation+0.075
Top resid features:
least
Token least
Feature activation+4.351
Top resid features:
they
Token they
Feature activation+0.373
Top resid features:
won
Token won
Feature activation+0.000
Top resid features:
't
Token't
Feature activation+0.000
Top resid features:
be
Token be
Feature activation+0.000
Top resid features:
exposed
Token exposed
Feature activation+0.000
Top resid features:
ling
Token ling
Feature activation+0.008
Top resid features:
ers
Tokeners
Feature activation+0.058
Top resid features:
,
Token,
Feature activation+0.050
Top resid features:
but
Token but
Feature activation+0.302
Top resid features:
at
Token at
Feature activation+0.147
Top resid features:
least
Token least
Feature activation+2.894
Top resid features:
there
Token there
Feature activation+0.913
Top resid features:
âĢ
TokenâĢ
Feature activation+0.111
Top resid features:
Ļ
TokenĻ
Feature activation-0.021
Top resid features:
s
Tokens
Feature activation+0.146
Top resid features:
been
Token been
Feature activation+0.000
Top resid features:
join
Token join
Feature activation-0.047
Top resid features:
the
Token the
Feature activation+0.011
Top resid features:
infantry
Token infantry
Feature activation-0.018
Top resid features:
,
Token,
Feature activation+0.032
Top resid features:
at
Token at
Feature activation+0.121
Top resid features:
least
Token least
Feature activation+4.065
Top resid features:
not
Token not
Feature activation+0.042
Top resid features:
immediately
Token immediately
Feature activation+0.000
Top resid features:
,
Token,
Feature activation+0.000
Top resid features:
since
Token since
Feature activation+0.000
Top resid features:
it
Token it
Feature activation+0.000
Top resid features:
play
Token play
Feature activation+0.011
Top resid features:
?
Token?
Feature activation+0.045
Top resid features:
Ċ
TokenĊ
Feature activation+0.045
Top resid features:
Ċ
TokenĊ
Feature activation+0.014
Top resid features:
At
TokenAt
Feature activation+0.083
Top resid features:
least
Token least
Feature activation+2.479
Top resid features:
they
Token they
Feature activation+1.402
Top resid features:
won
Token won
Feature activation+0.368
Top resid features:
't
Token't
Feature activation+0.327
Top resid features:
be
Token be
Feature activation+0.000
Top resid features:
exposed
Token exposed
Feature activation+0.000
Top resid features:
play
Token play
Feature activation+0.023
Top resid features:
?
Token?
Feature activation+0.009
Top resid features:
Ċ
TokenĊ
Feature activation+0.066
Top resid features:
Ċ
TokenĊ
Feature activation+0.046
Top resid features:
At
TokenAt
Feature activation+0.105
Top resid features:
least
Token least
Feature activation+2.002
Top resid features:
they
Token they
Feature activation+1.195
Top resid features:
won
Token won
Feature activation+0.350
Top resid features:
't
Token't
Feature activation+0.637
Top resid features:
be
Token be
Feature activation+0.190
Top resid features:
exposed
Token exposed
Feature activation+0.000
Top resid features:
been
Token been
Feature activation+0.007
Top resid features:
."
Token."
Feature activation+0.032
Top resid features:
Ċ
TokenĊ
Feature activation+0.048
Top resid features:
Ċ
TokenĊ
Feature activation+0.017
Top resid features:
At
TokenAt
Feature activation+0.089
Top resid features:
least
Token least
Feature activation+2.288
Top resid features:
the
Token the
Feature activation+0.440
Top resid features:
Lo
Token Lo
Feature activation-0.020
Top resid features:
ons
Tokenons
Feature activation+0.043
Top resid features:
won
Token won
Feature activation+0.406
Top resid features:
't
Token't
Feature activation+0.753
Top resid features:

Decoder Weights Distribution

Head 0: 0.05

Head 1: 0.04

Head 2: 0.01

Head 3: 0.08

Head 4: 0.03

Head 5: 0.09

Head 6: 0.05

Head 7: 0.02

Head 8: 0.10

Head 9: 0.50

Head 10: 0.02

Head 11: 0.03

Positive logits

appe1.91

usable1.69

spared1.67

ement1.67

acqu1.65

okay1.63

forgiving1.62

lasts1.61

retrospect1.61

sane1.60

apolog1.60

isse1.58

redeem1.54

usable1.53

prints1.52

rites1.52

avoids1.52

initialized1.52

ok1.51

honest1.51

Negative logits

INGTON-2.24

psey-2.16

GOODMAN-1.89

pecially-1.88

anwhile-1.88

okingly-1.84

Sabha-1.82

_-_-1.81

task-1.79

Brist-1.78

avascript-1.77

Dialogue-1.77

-1.76

GBT-1.75

adish-1.72

aciously-1.68

OTH-1.68

anguage-1.66

ELY-1.65

ESE-1.65

INTERVAL 5.357 - 5.952
CONTAINS 0.000%

least
Token least
Feature activation+0.937
I
Token I
Feature activation+4.081
didn
Token didn
Feature activation+3.322
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
t
Tokent
Feature activation+5.952
die
Token die
Feature activation+0.399
or
Token or
Feature activation+0.000
anything
Token anything
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000

INTERVAL 4.761 - 5.357
CONTAINS 0.000%

Hey
Token Hey
Feature activation+0.000
ladies
Token ladies
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.306
we
Token we
Feature activation+4.821
know
Token know
Feature activation+3.996
what
Token what
Feature activation+1.646
Donald
Token Donald
Feature activation+0.000
Trump
Token Trump
Feature activation+0.000
thinks
Token thinks
Feature activation+0.362
Ċ
TokenĊ
Feature activation+0.000
Well
TokenWell
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.451
I
Token I
Feature activation+4.829
drink
Token drink
Feature activation+1.974
beer
Token beer
Feature activation+0.852
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
a
Token a
Feature activation+0.000

INTERVAL 4.166 - 4.761
CONTAINS 0.000%

INTERVAL 3.571 - 4.166
CONTAINS 0.000%

ladies
Token ladies
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.306
we
Token we
Feature activation+4.821
know
Token know
Feature activation+3.996
what
Token what
Feature activation+1.646
Donald
Token Donald
Feature activation+0.000
Trump
Token Trump
Feature activation+0.000
thinks
Token thinks
Feature activation+0.362
of
Token of
Feature activation+0.000
back
Token back
Feature activation+0.000
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.937
I
Token I
Feature activation+4.081
didn
Token didn
Feature activation+3.322
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
t
Tokent
Feature activation+5.952
die
Token die
Feature activation+0.399
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.993
she
Token she
Feature activation+3.099
's
Token's
Feature activation+3.692
slightly
Token slightly
Feature activation+1.975
more
Token more
Feature activation+1.462
rel
Token rel
Feature activation+0.000
atable
Tokenatable
Feature activation+0.997
.
Token.
Feature activation+0.000
in
Token in
Feature activation+0.000
Brazil
Token Brazil
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+3.838
will
Token will
Feature activation+3.301
return
Token return
Feature activation+0.522
home
Token home
Feature activation+0.621
with
Token with
Feature activation+0.037
clean
Token clean
Feature activation+0.000

INTERVAL 2.976 - 3.571
CONTAINS 0.001%

Brazil
Token Brazil
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+3.838
will
Token will
Feature activation+3.301
return
Token return
Feature activation+0.522
home
Token home
Feature activation+0.621
with
Token with
Feature activation+0.037
clean
Token clean
Feature activation+0.000
cons
Token cons
Feature activation+0.000
At
Token At
Feature activation+0.000
least
Token least
Feature activation+0.000
for
Token for
Feature activation+0.523
me
Token me
Feature activation+0.784
,
Token,
Feature activation+1.456
I
Token I
Feature activation+3.272
mean
Token mean
Feature activation+0.599
,
Token,
Feature activation+1.342
shit
Token shit
Feature activation+0.424
,
Token,
Feature activation+0.707
I
Token I
Feature activation+0.574
world
Token world
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
the
Token the
Feature activation+0.166
world
Token world
Feature activation+3.062
inside
Token inside
Feature activation+1.155
the
Token the
Feature activation+0.000
home
Token home
Feature activation+0.000
,
Token,
Feature activation+0.000
I
Token I
Feature activation+0.000
least
Token least
Feature activation+0.000
the
Token the
Feature activation+1.859
Lo
Token Lo
Feature activation+0.000
ons
Tokenons
Feature activation+0.853
won
Token won
Feature activation+2.606
't
Token't
Feature activation+3.426
be
Token be
Feature activation+2.775
missing
Token missing
Feature activation+1.577
any
Token any
Feature activation+0.794
key
Token key
Feature activation+0.307
pieces
Token pieces
Feature activation+0.895
AI
Token AI
Feature activation+0.000
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.993
she
Token she
Feature activation+3.099
's
Token's
Feature activation+3.692
slightly
Token slightly
Feature activation+1.975
more
Token more
Feature activation+1.462
rel
Token rel
Feature activation+0.000
atable
Tokenatable
Feature activation+0.997

INTERVAL 2.381 - 2.976
CONTAINS 0.001%

At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+2.932
won
Token won
Feature activation+1.973
't
Token't
Feature activation+2.789
be
Token be
Feature activation+2.789
exposed
Token exposed
Feature activation+0.667
to
Token to
Feature activation+0.518
the
Token the
Feature activation+0.000
"
Token "
Feature activation+0.000
extreme
Tokenextreme
Feature activation+0.000
?
Token?
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+2.932
won
Token won
Feature activation+1.973
't
Token't
Feature activation+2.789
be
Token be
Feature activation+2.789
exposed
Token exposed
Feature activation+0.667
to
Token to
Feature activation+0.518
the
Token the
Feature activation+1.859
Lo
Token Lo
Feature activation+0.000
ons
Tokenons
Feature activation+0.853
won
Token won
Feature activation+2.606
't
Token't
Feature activation+3.426
be
Token be
Feature activation+2.775
missing
Token missing
Feature activation+1.577
any
Token any
Feature activation+0.794
key
Token key
Feature activation+0.307
pieces
Token pieces
Feature activation+0.895
heading
Token heading
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
the
Token the
Feature activation+1.859
Lo
Token Lo
Feature activation+0.000
ons
Tokenons
Feature activation+0.853
won
Token won
Feature activation+2.606
't
Token't
Feature activation+3.426
be
Token be
Feature activation+2.775
missing
Token missing
Feature activation+1.577
any
Token any
Feature activation+0.794
key
Token key
Feature activation+0.307
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.846
there
Token there
Feature activation+2.243
âĢ
TokenâĢ
Feature activation+0.000
Ļ
TokenĻ
Feature activation+0.000
s
Tokens
Feature activation+2.812
been
Token been
Feature activation+1.253
improvement
Token improvement
Feature activation+0.129
,
Token,
Feature activation+0.443
said
Token said
Feature activation+0.000
David
Token David
Feature activation+0.000

INTERVAL 1.786 - 2.381
CONTAINS 0.001%

Well
TokenWell
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.451
I
Token I
Feature activation+4.829
drink
Token drink
Feature activation+1.974
beer
Token beer
Feature activation+0.852
.
Token.
Feature activation+0.000
As
Token As
Feature activation+0.000
a
Token a
Feature activation+0.000
beer
Token beer
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
,
Token,
Feature activation+0.358
you
Token you
Feature activation+1.906
can
Token can
Feature activation+1.414
in
Token in
Feature activation+0.914
October
Token October
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
."
Token."
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
the
Token the
Feature activation+1.859
Lo
Token Lo
Feature activation+0.000
ons
Tokenons
Feature activation+0.853
won
Token won
Feature activation+2.606
't
Token't
Feature activation+3.426
be
Token be
Feature activation+2.775
song
Token song
Feature activation+0.000
,
Token,
Feature activation+0.000
so
Token so
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.318
someone
Token someone
Feature activation+2.259
who
Token who
Feature activation+0.264
was
Token was
Feature activation+0.599
interested
Token interested
Feature activation+0.000
in
Token in
Feature activation+0.000
the
Token the
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
At
TokenAt
Feature activation+0.000
least
Token least
Feature activation+0.000
they
Token they
Feature activation+2.932
won
Token won
Feature activation+1.973
't
Token't
Feature activation+2.789
be
Token be
Feature activation+2.789
exposed
Token exposed
Feature activation+0.667
to
Token to
Feature activation+0.518
the
Token the
Feature activation+0.000

INTERVAL 1.190 - 1.786
CONTAINS 0.002%

,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.306
we
Token we
Feature activation+4.821
know
Token know
Feature activation+3.996
what
Token what
Feature activation+1.646
Donald
Token Donald
Feature activation+0.000
Trump
Token Trump
Feature activation+0.000
thinks
Token thinks
Feature activation+0.362
of
Token of
Feature activation+0.000
us
Token us
Feature activation+0.000
House
Token House
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
,
Token,
Feature activation+1.056
is
Token is
Feature activation+1.265
big
Token big
Feature activation+0.000
enough
Token enough
Feature activation+0.000
to
Token to
Feature activation+0.000
do
Token do
Feature activation+0.000
whatever
Token whatever
Feature activation+0.000
Lo
Token Lo
Feature activation+0.000
ons
Tokenons
Feature activation+0.853
won
Token won
Feature activation+2.606
't
Token't
Feature activation+3.426
be
Token be
Feature activation+2.775
missing
Token missing
Feature activation+1.577
any
Token any
Feature activation+0.794
key
Token key
Feature activation+0.307
pieces
Token pieces
Feature activation+0.895
heading
Token heading
Feature activation+0.000
into
Token into
Feature activation+0.000
join
Token join
Feature activation+0.000
the
Token the
Feature activation+0.000
infantry
Token infantry
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+1.288
not
Token not
Feature activation+2.812
immediately
Token immediately
Feature activation+0.533
,
Token,
Feature activation+0.000
since
Token since
Feature activation+0.000
it
Token it
Feature activation+0.000
you
Token you
Feature activation+0.000
can
Token can
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.051
use
Token use
Feature activation+0.080
them
Token them
Feature activation+1.208
to
Token to
Feature activation+0.061
question
Token question
Feature activation+1.115
your
Token your
Feature activation+0.000
evaluation
Token evaluation
Feature activation+0.000
that
Token that
Feature activation+0.000

INTERVAL 0.595 - 1.190
CONTAINS 0.004%

,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
she
Token she
Feature activation+1.647
was
Token was
Feature activation+1.071
seen
Token seen
Feature activation+1.141
as
Token as
Feature activation+0.771
cleaning
Token cleaning
Feature activation+0.000
up
Token up
Feature activation+0.000
after
Token after
Feature activation+0.000
earlier
Token earlier
Feature activation+0.000
or
Tokenor
Feature activation+0.000
,
Token,
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.000
,
Token,
Feature activation+0.432
it
Token it
Feature activation+0.945
should
Token should
Feature activation+0.297
be
Token be
Feature activation+0.661
!
Token!
Feature activation+0.000
There
Token There
Feature activation+0.000
might
Token might
Feature activation+0.000
least
Token least
Feature activation+1.320
they
Token they
Feature activation+3.260
could
Token could
Feature activation+2.459
allocate
Token allocate
Feature activation+0.486
some
Token some
Feature activation+0.343
profits
Token profits
Feature activation+0.816
to
Token to
Feature activation+1.088
keep
Token keep
Feature activation+0.649
the
Token the
Feature activation+0.159
rest
Token rest
Feature activation+0.288
of
Token of
Feature activation+0.000
of
Token of
Feature activation+0.000
AI
Token AI
Feature activation+0.000
,
Token,
Feature activation+0.000
but
Token but
Feature activation+0.000
at
Token at
Feature activation+0.000
least
Token least
Feature activation+0.993
she
Token she
Feature activation+3.099
's
Token's
Feature activation+3.692
slightly
Token slightly
Feature activation+1.975
more
Token more
Feature activation+1.462
rel
Token rel
Feature activation+0.000
could
Token could
Feature activation+2.459
allocate
Token allocate
Feature activation+0.486
some
Token some
Feature activation+0.343
profits
Token profits
Feature activation+0.816
to
Token to
Feature activation+1.088
keep
Token keep
Feature activation+0.649
the
Token the
Feature activation+0.159
rest
Token rest
Feature activation+0.288
of
Token of
Feature activation+0.000
us
Token us
Feature activation+0.038
alive
Token alive
Feature activation+0.000

INTERVAL 0.000 - 0.595
CONTAINS 99.991%

at
Token at
Feature activation+0.000
the
Token the
Feature activation+0.000
public
Token public
Feature activation+0.000
meeting
Token meeting
Feature activation+0.000
last
Token last
Feature activation+0.000
night
Token night
Feature activation+0.000
,
Token,
Feature activation+0.000
said
Token said
Feature activation+0.000
both
Token both
Feature activation+0.000
Dublin
Token Dublin
Feature activation+0.000
City
Token City
Feature activation+0.000
âĢ
Token âĢ
Feature activation+0.000
ľ
Tokenľ
Feature activation+0.000
really
Tokenreally
Feature activation+0.000
,
Token,
Feature activation+0.000
really
Token really
Feature activation+0.000
good
Token good
Feature activation+0.000
senator
Token senator
Feature activation+0.000
.
Token.
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
ig
Token ig
Feature activation+0.000
lo
Tokenlo
Feature activation+0.000
os
Tokenos
Feature activation+0.000
.[
Token.[
Feature activation+0.000
23
Token23
Feature activation+0.000
]
Token]
Feature activation+0.000
Spe
Token Spe
Feature activation+0.000
culation
Tokenculation
Feature activation+0.000
on
Token on
Feature activation+0.000
the
Token the
Feature activation+0.000
tunnels
Token tunnels
Feature activation+0.000
--
Token --
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Former
TokenFormer
Feature activation+0.000
Mayor
Token Mayor
Feature activation+0.000
Lee
Token Lee
Feature activation+0.000
Le
Token Le
Feature activation+0.000
ff
Tokenff
Feature activation+0.000
ing
Tokening
Feature activation+0.000
well
Tokenwell
Feature activation+0.000
knows
Token knows
Feature activation+0.000
even
Token even
Feature activation+0.000
a
Token a
Feature activation+0.000
breastfeeding
Token breastfeeding
Feature activation+0.000
advocate
Token advocate
Feature activation+0.000
like
Token like
Feature activation+0.000
myself
Token myself
Feature activation+0.000
.
Token.
Feature activation+0.000
âĢ
TokenâĢ
Feature activation+0.000
Ŀ
TokenĿ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000

BOTTOM ACTIVATIONS
MIN = 0.000

atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
Development
Token Development
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
&
Token &
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Defence
Token Defence
Feature activation+0.000
Research
Token Research
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
<|endoftext|>
Token<|endoftext|>
Feature activation+0.000
rik
Tokenrik
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
ar
Tokenar
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
said
Token said
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
A
Token A
Feature activation+0.000
atre
Tokenatre
Feature activation+0.000
,
Token,
Feature activation+0.000
a
Token a
Feature activation+0.000
former
Token former
Feature activation+0.000
boss
Token boss
Feature activation+0.000
of
Token of
Feature activation+0.000
the
Token the
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
K
TokenK
Feature activation+0.000
Saturday
Token Saturday
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
defence
Token defence
Feature activation+0.000
ministry
Token ministry
Feature activation+0.000
committee
Token committee
Feature activation+0.000
headed
Token headed
Feature activation+0.000
by
Token by
Feature activation+0.000
V
Token V
Feature activation+0.000
.
Token.
Feature activation+0.000

Top feature 5 in H7.9: (feature 29342)

TOP ACTIVATIONS
MAX = 2.499

Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
government
Token government
Feature activation+0.000
felt
Token felt
Feature activation+0.000
children
Token children
Feature activation+1.812
were
Token were
Feature activation+2.499
easier
Token easier
Feature activation+0.245
to
Token to
Feature activation+0.295
mould
Token mould
Feature activation+0.000
than
Token than
Feature activation+0.000
adults
Token adults
Feature activation+0.000
Hugh
Token Hugh
Feature activation+0.000
Glass
Token Glass
Feature activation+0.000
because
Token because
Feature activation+0.000
she
Token she
Feature activation+0.000
feels
Token feels
Feature activation+0.000
he
Token he
Feature activation+2.376
might
Token might
Feature activation+0.794
be
Token be
Feature activation+0.770
threatening
Token threatening
Feature activation+0.632
her
Token her
Feature activation+0.508
cub
Token cub
Feature activation+0.000
âĢĶ
TokenâĢĶ
Feature activation+0.145
and
Tokenand
Feature activation+0.000
yes
Token yes
Feature activation+0.000
women
Token women
Feature activation+0.000
âĢĶ
TokenâĢĶ
Feature activation+0.787
would
Tokenwould
Feature activation+2.094
appreciate
Token appreciate
Feature activation+0.493
a
Token a
Feature activation+0.269
"
Token "
Feature activation+0.000
sc
Tokensc
Feature activation+0.000
rup
Tokenrup
Feature activation+0.000
,
Token,
Feature activation+0.000
I
Token I
Feature activation+0.000
felt
Token felt
Feature activation+0.000
that
Token that
Feature activation+0.498
wasn
Token wasn
Feature activation+1.629
't
Token't
Feature activation+2.038
creative
Token creative
Feature activation+0.371
enough
Token enough
Feature activation+0.437
for
Token for
Feature activation+0.554
me
Token me
Feature activation+0.000
.
Token.
Feature activation+0.000
.
Token.
Feature activation+0.000
We
Token We
Feature activation+0.000
believe
Token believe
Feature activation+0.000
that
Token that
Feature activation+0.437
it
Token it
Feature activation+1.468
is
Token is
Feature activation+1.972
an
Token an
Feature activation+1.823
ideal
Token ideal
Feature activation+0.558
place
Token place
Feature activation+1.074
to
Token to
Feature activation+1.066
show
Token show
Feature activation+0.298
NSA
Token NSA
Feature activation+0.000
,
Token,
Feature activation+0.000
thinks
Token thinks
Feature activation+0.000
the
Token the
Feature activation+0.111
US
Token US
Feature activation+0.657
government
Token government
Feature activation+1.929
should
Token should
Feature activation+0.509
stop
Token stop
Feature activation+0.000
railing
Token railing
Feature activation+0.000
against
Token against
Feature activation+0.000
encryption
Token encryption
Feature activation+0.000
power
Token power
Feature activation+0.000
couple
Token couple
Feature activation+0.000
because
Token because
Feature activation+0.000
he
Token he
Feature activation+0.000
believes
Token believes
Feature activation+0.125
they
Token they
Feature activation+1.926
lied
Token lied
Feature activation+0.372
'
Token '
Feature activation+0.130
with
Tokenwith
Feature activation+0.056
such
Token such
Feature activation+0.000
ease
Token ease
Feature activation+0.051
not
Token not
Feature activation+0.000
believe
Token believe
Feature activation+0.000
that
Token that
Feature activation+0.028
the
Token the
Feature activation+0.323
coin
Token coin
Feature activation+0.222
is
Token is
Feature activation+1.893
worth
Token worth
Feature activation+1.296
$
Token $
Feature activation+0.000
1
Token1
Feature activation+0.000
trillion
Token trillion
Feature activation+0.602
and
Token and
Feature activation+0.337
felt
Token felt
Feature activation+0.000
that
Token that
Feature activation+0.000
H
Token H
Feature activation+0.000
aged
Tokenaged
Feature activation+0.000
orn
Tokenorn
Feature activation+1.660
was
Token was
Feature activation+1.852
qualified
Token qualified
Feature activation+0.172
for
Token for
Feature activation+0.311
the
Token the
Feature activation+0.000
things
Token things
Feature activation+0.000
he
Token he
Feature activation+0.000
We
Token We
Feature activation+0.000
believe
Token believe
Feature activation+0.000
that
Token that
Feature activation+0.437
it
Token it
Feature activation+1.468
is
Token is
Feature activation+1.972
an
Token an
Feature activation+1.823
ideal
Token ideal
Feature activation+0.558
place
Token place
Feature activation+1.074
to
Token to
Feature activation+1.066
show
Token show
Feature activation+0.298
the
Token the
Feature activation+0.000
ire
Tokenire
Feature activation+0.000
children
Token children
Feature activation+0.000
who
Token who
Feature activation+0.000
believe
Token believe
Feature activation+0.000
a
Token a
Feature activation+1.096
criminal
Token criminal
Feature activation+1.822
they
Token they
Feature activation+0.501
discover
Token discover
Feature activation+0.000
hiding
Token hiding
Feature activation+0.916
in
Token in
Feature activation+0.000
their
Token their
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
government
Token government
Feature activation+0.000
felt
Token felt
Feature activation+0.000
children
Token children
Feature activation+1.812
were
Token were
Feature activation+2.499
easier
Token easier
Feature activation+0.245
to
Token to
Feature activation+0.295
mould
Token mould
Feature activation+0.000
than
Token than
Feature activation+0.000
.
Token.
Feature activation+0.000
Bel
Token Bel
Feature activation+0.000
ieving
Tokenieving
Feature activation+0.341
the
Token the
Feature activation+0.527
suspect
Token suspect
Feature activation+1.668
was
Token was
Feature activation+1.775
armed
Token armed
Feature activation+0.183
,
Token,
Feature activation+0.000
"
Token "
Feature activation+0.000
both
Tokenboth
Feature activation+0.000
officers
Token officers
Feature activation+0.000
Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
woman
Token woman
Feature activation+0.000
feels
Token feels
Feature activation+0.000
it
Token it
Feature activation+1.528
takes
Token takes
Feature activation+1.711
an
Token an
Feature activation+0.538
entire
Token entire
Feature activation+0.000
community
Token community
Feature activation+0.189
to
Token to
Feature activation+0.080
solve
Token solve
Feature activation+0.000
who
Token who
Feature activation+0.000
turned
Token turned
Feature activation+0.000
28
Token 28
Feature activation+0.000
on
Token on
Feature activation+0.000
Tuesday
Token Tuesday
Feature activation+0.000
,
Token,
Feature activation+1.702
can
Token can
Feature activation+0.977
be
Token be
Feature activation+1.193
a
Token a
Feature activation+0.746
stead
Token stead
Feature activation+0.000
ying
Tokenying
Feature activation+0.000
Matt
Token Matt
Feature activation+0.000
and
Token and
Feature activation+0.000
Dem
Token Dem
Feature activation+0.000
enc
Tokenenc
Feature activation+0.000
io
Tokenio
Feature activation+0.000
are
Token are
Feature activation+1.701
related
Token related
Feature activation+0.000
,
Token,
Feature activation+0.000
it
Token it
Feature activation+0.000
is
Token is
Feature activation+0.000
a
Token a
Feature activation+0.000
"
Token"
Feature activation+0.000
We
TokenWe
Feature activation+0.000
believe
Token believe
Feature activation+0.000
Med
Token Med
Feature activation+0.000
ved
Tokenved
Feature activation+0.000
ev
Tokenev
Feature activation+1.689
and
Token and
Feature activation+0.000
Putin
Token Putin
Feature activation+0.820
,
Token,
Feature activation+0.000
as
Token as
Feature activation+0.000
prime
Token prime
Feature activation+0.000
Department
Token Department
Feature activation+0.000
.
Token.
Feature activation+0.000
Bel
Token Bel
Feature activation+0.000
ieving
Tokenieving
Feature activation+0.341
the
Token the
Feature activation+0.527
suspect
Token suspect
Feature activation+1.668
was
Token was
Feature activation+1.775
armed
Token armed
Feature activation+0.183
,
Token,
Feature activation+0.000
"
Token "
Feature activation+0.000
both
Tokenboth
Feature activation+0.000
Walker
Token Walker
Feature activation+0.000
felt
Token felt
Feature activation+0.000
that
Token that
Feature activation+0.000
H
Token H
Feature activation+0.000
aged
Tokenaged
Feature activation+0.000
orn
Tokenorn
Feature activation+1.660
was
Token was
Feature activation+1.852
qualified
Token qualified
Feature activation+0.172
for
Token for
Feature activation+0.311
the
Token the
Feature activation+0.000
things
Token things
Feature activation+0.000
particular
Token particular
Feature activation+0.000
issue
Token issue
Feature activation+0.000
,
Token,
Feature activation+0.000
we
Token we
Feature activation+0.000
believe
Token believe
Feature activation+0.000
,
Token,
Feature activation+1.658
is
Token is
Feature activation+0.990
a
Token a
Feature activation+0.320
red
Token red
Feature activation+0.000
her
Token her
Feature activation+0.000
ring
Tokenring
Feature activation+0.088

Top DFA by src position
MAX = 3.071

.
Token.
Feature activation+0.034
Top resid features:
Ċ
TokenĊ
Feature activation-0.000
Top resid features:
Ċ
TokenĊ
Feature activation+0.001
Top resid features:
The
TokenThe
Feature activation-0.003
Top resid features:
government
Token government
Feature activation+0.006
Top resid features:
felt
Token felt
Feature activation+3.071
Top resid features:
children
Token children
Feature activation+0.128
Top resid features:
were
Token were
Feature activation+0.028
Top resid features:
easier
Token easier
Feature activation+0.000
Top resid features:
to
Token to
Feature activation+0.000
Top resid features:
mould
Token mould
Feature activation+0.000
Top resid features:
attacks
Token attacks
Feature activation+0.158
Top resid features:
Hugh
Token Hugh
Feature activation-0.033
Top resid features:
Glass
Token Glass
Feature activation+0.053
Top resid features:
because
Token because
Feature activation+0.044
Top resid features:
she
Token she
Feature activation+0.034
Top resid features:
feels
Token feels
Feature activation+2.821
Top resid features:
he
Token he
Feature activation+0.014
Top resid features:
might
Token might
Feature activation+0.000
Top resid features:
be
Token be
Feature activation+0.000
Top resid features:
threatening
Token threatening
Feature activation+0.000
Top resid features:
her
Token her
Feature activation+0.000
Top resid features:
them
Token them
Feature activation+0.004
Top resid features:
.
Token.
Feature activation+0.004
Top resid features:
Ċ
TokenĊ
Feature activation-0.005
Top resid features:
Ċ
TokenĊ
Feature activation-0.013
Top resid features:
Bel
TokenBel
Feature activation+0.008
Top resid features:
ieving
Tokenieving
Feature activation+2.752
Top resid features:
men
Token men
Feature activation+0.018
Top resid features:
âĢĶ
TokenâĢĶ
Feature activation+0.007
Top resid features:
and
Tokenand
Feature activation+0.005
Top resid features:
yes
Token yes
Feature activation-0.004
Top resid features:
women
Token women
Feature activation+0.023
Top resid features:
with
Token with
Feature activation-0.004
Top resid features:
waste
Token waste
Feature activation-0.005
Top resid features:
not
Token not
Feature activation-0.008
Top resid features:
,
Token,
Feature activation-0.009
Top resid features:
I
Token I
Feature activation+0.048
Top resid features:
felt
Token felt
Feature activation+2.460
Top resid features:
that
Token that
Feature activation+0.308
Top resid features:
wasn
Token wasn
Feature activation+0.069
Top resid features:
't
Token't
Feature activation-0.006
Top resid features:
creative
Token creative
Feature activation+0.000
Top resid features:
enough
Token enough
Feature activation+0.000
Top resid features:
log
Token log
Feature activation-0.002
Top resid features:
istic
Tokenistic
Feature activation-0.000
Top resid features:
center
Token center
Feature activation+0.014
Top resid features:
.
Token.
Feature activation-0.012
Top resid features:
We
Token We
Feature activation+0.018
Top resid features:
believe
Token believe
Feature activation+1.951
Top resid features:
that
Token that
Feature activation+0.459
Top resid features:
it
Token it
Feature activation+0.103
Top resid features:
is
Token is
Feature activation+0.036
Top resid features:
an
Token an
Feature activation+0.000
Top resid features:
ideal
Token ideal
Feature activation+0.000
Top resid features:
,
Token,
Feature activation+0.003
Top resid features:
and
Token and
Feature activation+0.001
Top resid features:
the
Token the
Feature activation+0.005
Top resid features:
NSA
Token NSA
Feature activation+0.004
Top resid features:
,
Token,
Feature activation+0.008
Top resid features:
thinks
Token thinks
Feature activation+2.427
Top resid features:
the
Token the
Feature activation+0.116
Top resid features:
US
Token US
Feature activation+0.007
Top resid features:
government
Token government
Feature activation-0.022
Top resid features:
should
Token should
Feature activation+0.000
Top resid features:
stop
Token stop
Feature activation+0.000
Top resid features:
political
Token political
Feature activation-0.004
Top resid features:
power
Token power
Feature activation-0.002
Top resid features:
couple
Token couple
Feature activation-0.004
Top resid features:
because
Token because
Feature activation+0.100
Top resid features:
he
Token he
Feature activation+0.044
Top resid features:
believes
Token believes
Feature activation+2.268
Top resid features:
they
Token they
Feature activation+0.046
Top resid features:
lied
Token lied
Feature activation+0.000
Top resid features:
'
Token '
Feature activation+0.000
Top resid features:
with
Tokenwith
Feature activation+0.000
Top resid features:
such
Token such
Feature activation+0.000
Top resid features:
the
Token the
Feature activation+0.004
Top resid features:
Federal
Token Federal
Feature activation+0.001
Top resid features:
Reserve
Token Reserve
Feature activation+0.007
Top resid features:
does
Token does
Feature activation+0.042
Top resid features:
not
Token not
Feature activation+0.023
Top resid features:
believe
Token believe
Feature activation+1.888
Top resid features:
that
Token that
Feature activation+0.387
Top resid features:
the
Token the
Feature activation+0.097
Top resid features:
coin
Token coin
Feature activation+0.125
Top resid features:
is
Token is
Feature activation+0.088
Top resid features:
worth
Token worth
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.005
Top resid features:
Ċ
TokenĊ
Feature activation-0.002
Top resid features:
Ċ
TokenĊ
Feature activation-0.004
Top resid features:
Perhaps
TokenPerhaps
Feature activation+0.064
Top resid features:
Walker
Token Walker
Feature activation-0.021
Top resid features:
felt
Token felt
Feature activation+2.131
Top resid features:
that
Token that
Feature activation+0.394
Top resid features:
H
Token H
Feature activation-0.024
Top resid features:
aged
Tokenaged
Feature activation-0.002
Top resid features:
orn
Tokenorn
Feature activation+0.002
Top resid features:
was
Token was
Feature activation+0.018
Top resid features:
log
Token log
Feature activation-0.002
Top resid features:
istic
Tokenistic
Feature activation-0.001
Top resid features:
center
Token center
Feature activation+0.033
Top resid features:
.
Token.
Feature activation-0.011
Top resid features:
We
Token We
Feature activation+0.012
Top resid features:
believe
Token believe
Feature activation+1.655
Top resid features:
that
Token that
Feature activation+0.444
Top resid features:
it
Token it
Feature activation+0.105
Top resid features:
is
Token is
Feature activation+0.124
Top resid features:
an
Token an
Feature activation+0.029
Top resid features:
ideal
Token ideal
Feature activation+0.000
Top resid features:
Lanc
Token Lanc
Feature activation-0.005
Top resid features:
ash
Tokenash
Feature activation-0.002
Top resid features:
ire
Tokenire
Feature activation-0.002
Top resid features:
children
Token children
Feature activation-0.000
Top resid features:
who
Token who
Feature activation+0.033
Top resid features:
believe
Token believe
Feature activation+2.164
Top resid features:
a
Token a
Feature activation+0.353
Top resid features:
criminal
Token criminal
Feature activation+0.011
Top resid features:
they
Token they
Feature activation+0.000
Top resid features:
discover
Token discover
Feature activation+0.000
Top resid features:
hiding
Token hiding
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.030
Top resid features:
Ċ
TokenĊ
Feature activation-0.003
Top resid features:
Ċ
TokenĊ
Feature activation+0.001
Top resid features:
The
TokenThe
Feature activation-0.004
Top resid features:
government
Token government
Feature activation+0.004
Top resid features:
felt
Token felt
Feature activation+2.518
Top resid features:
children
Token children
Feature activation+0.009
Top resid features:
were
Token were
Feature activation+0.000
Top resid features:
easier
Token easier
Feature activation+0.000
Top resid features:
to
Token to
Feature activation+0.000
Top resid features:
mould
Token mould
Feature activation+0.000
Top resid features:
Jose
Token Jose
Feature activation-0.000
Top resid features:
Police
Token Police
Feature activation+0.000
Top resid features:
Department
Token Department
Feature activation-0.000
Top resid features:
.
Token.
Feature activation-0.014
Top resid features:
Bel
Token Bel
Feature activation+0.049
Top resid features:
ieving
Tokenieving
Feature activation+1.850
Top resid features:
the
Token the
Feature activation+0.139
Top resid features:
suspect
Token suspect
Feature activation+0.354
Top resid features:
was
Token was
Feature activation+0.059
Top resid features:
armed
Token armed
Feature activation+0.000
Top resid features:
,
Token,
Feature activation+0.000
Top resid features:
.
Token.
Feature activation-0.012
Top resid features:
Ċ
TokenĊ
Feature activation-0.012
Top resid features:
Ċ
TokenĊ
Feature activation-0.015
Top resid features:
The
TokenThe
Feature activation-0.012
Top resid features:
woman
Token woman
Feature activation+0.052
Top resid features:
feels
Token feels
Feature activation+2.163
Top resid features:
it
Token it
Feature activation+0.170
Top resid features:
takes
Token takes
Feature activation-0.005
Top resid features:
an
Token an
Feature activation+0.000
Top resid features:
entire
Token entire
Feature activation+0.000
Top resid features:
community
Token community
Feature activation+0.000
Top resid features:
Ċ
TokenĊ
Feature activation-0.009
Top resid features:
Ċ
TokenĊ
Feature activation-0.012
Top resid features:
GM
TokenGM
Feature activation-0.000
Top resid features:
Ryan
Token Ryan
Feature activation-0.009
Top resid features:
Pace
Token Pace
Feature activation+0.007
Top resid features:
believes
Token believes
Feature activation+2.169
Top resid features:
Am
Token Am
Feature activation+0.036
Top resid features:
uk
Tokenuk
Feature activation+0.031
Top resid features:
amara
Tokenamara
Feature activation+0.217
Top resid features:
,
Token,
Feature activation+0.137
Top resid features:
who
Token who
Feature activation-0.003
Top resid features:
ľ
Tokenľ
Feature activation+0.009
Top resid features:
While
TokenWhile
Feature activation-0.026
Top resid features:
I
Token I
Feature activation+0.003
Top resid features:
do
Token do
Feature activation+0.054
Top resid features:
not
Token not
Feature activation+0.066
Top resid features:
believe
Token believe
Feature activation+2.163
Top resid features:
the
Token the
Feature activation+0.150
Top resid features:
murders
Token murders
Feature activation-0.093
Top resid features:
of
Token of
Feature activation-0.018
Top resid features:
Matt
Token Matt
Feature activation-0.008
Top resid features:
and
Token and
Feature activation+0.006
Top resid features:
.
Token.
Feature activation+0.002
Top resid features:
Ċ
TokenĊ
Feature activation+0.001
Top resid features:
Ċ
TokenĊ
Feature activation+0.001
Top resid features:
"
Token"
Feature activation-0.061
Top resid features:
We
TokenWe
Feature activation-0.013
Top resid features:
believe
Token believe
Feature activation+2.545
Top resid features:
Med
Token Med
Feature activation+0.173
Top resid features:
ved
Tokenved
Feature activation-0.054
Top resid features:
ev
Tokenev
Feature activation-0.052
Top resid features:
and
Token and
Feature activation+0.000
Top resid features:
Putin
Token Putin
Feature activation+0.000
Top resid features:
Jose
Token Jose
Feature activation-0.000
Top resid features:
Police
Token Police
Feature activation-0.000
Top resid features:
Department
Token Department
Feature activation-0.001
Top resid features:
.
Token.
Feature activation-0.037
Top resid features:
Bel
Token Bel
Feature activation+0.037
Top resid features:
ieving
Tokenieving
Feature activation+1.915
Top resid features:
the
Token the
Feature activation+0.293
Top resid features:
suspect
Token suspect
Feature activation+0.111
Top resid features:
was
Token was
Feature activation+0.000
Top resid features:
armed
Token armed
Feature activation+0.000
Top resid features:
,
Token,
Feature activation+0.000
Top resid features:
.
Token.
Feature activation+0.007
Top resid features:
Ċ
TokenĊ
Feature activation+0.006
Top resid features:
Ċ
TokenĊ
Feature activation-0.001
Top resid features:
Perhaps
TokenPerhaps
Feature activation+0.071
Top resid features:
Walker
Token Walker
Feature activation-0.006
Top resid features:
felt
Token felt
Feature activation+1.560
Top resid features:
that
Token that
Feature activation+0.884
Top resid features:
H
Token H
Feature activation+0.025
Top resid features:
aged
Tokenaged
Feature activation+0.003
Top resid features:
orn
Tokenorn
Feature activation+0.018
Top resid features:
was
Token was
Feature activation+0.000
Top resid features:
This
Token This
Feature activation-0.013
Top resid features:
particular
Token particular
Feature activation+0.004
Top resid features:
issue
Token issue
Feature activation-0.169
Top resid features:
,
Token,
Feature activation-0.041
Top resid features:
we
Token we
Feature activation-0.115
Top resid features:
believe
Token believe
Feature activation+2.623
Top resid features:
,
Token,
Feature activation+0.077
Top resid features:
is
Token is
Feature activation+0.000
Top resid features:
a
Token a
Feature activation+0.000
Top resid features:
red
Token red
Feature activation+0.000
Top resid features:
her
Token her
Feature activation+0.000
Top resid features:

Decoder Weights Distribution

Head 0: 0.06

Head 1: 0.04

Head 2: 0.01

Head 3: 0.12

Head 4: 0.05

Head 5: 0.07

Head 6: 0.04

Head 7: 0.02

Head 8: 0.06

Head 9: 0.49

Head 10: 0.02

Head 11: 0.02

Positive logits

ought2.46

deserved2.42

deserves2.40

unfairly2.37

deserve2.33

should2.23

outweigh2.04

adequately1.99

sufficiently1.98

adequate1.97

wiser1.93

unfair1.92

superior1.90

cheat1.90

cible1.87

sufficient1.87

preferable1.86

better1.85

entit1.84

healthier1.81

Negative logits

andan-2.03

ierrez-2.03

ggles-2.02

obin-2.01

toggle-2.00

FTWARE-1.96

wondered-1.94

checking-1.92

anwhile-1.89

livious-1.84

umbnails-1.82

arel-1.80

ゴン-1.76

braces-1.75

ilight-1.70

yrinth-1.70

mares-1.70

asma-1.68

idav-1.68

quartered-1.66

INTERVAL 2.249 - 2.499
CONTAINS 0.000%

Ċ
TokenĊ
Feature activation+0.000
The
TokenThe
Feature activation+0.000
government
Token government
Feature activation+0.000
felt
Token felt
Feature activation+0.000
children
Token children
Feature activation+1.812
were
Token were
Feature activation+2.499
easier
Token easier
Feature activation+0.245
to
Token to
Feature activation+0.295
mould
Token mould
Feature activation+0.000
than
Token than
Feature activation+0.000
adults
Token adults
Feature activation+0.000
Hugh
Token Hugh
Feature activation+0.000
Glass
Token Glass
Feature activation+0.000
because
Token because
Feature activation+0.000
she
Token she
Feature activation+0.000
feels
Token feels
Feature activation+0.000
he
Token he
Feature activation+2.376
might
Token might
Feature activation+0.794
be
Token be
Feature activation+0.770
threatening
Token threatening
Feature activation+0.632
her
Token her
Feature activation+0.508
cub
Token cub
Feature activation+0.000

INTERVAL 1.999 - 2.249
CONTAINS 0.000%

âĢĶ
TokenâĢĶ
Feature activation+0.145
and
Tokenand
Feature activation+0.000
yes
Token yes
Feature activation+0.000
women
Token women
Feature activation+0.000
âĢĶ
TokenâĢĶ
Feature activation+0.787
would
Tokenwould
Feature activation+2.094
appreciate
Token appreciate
Feature activation+0.493
a
Token a
Feature activation+0.269
"
Token "
Feature activation+0.000
sc
Tokensc
Feature activation+0.000
rup
Tokenrup
Feature activation+0.000
,
Token,
Feature activation+0.000
I
Token I
Feature activation+0.000
felt
Token felt
Feature activation+0.000
that
Token that
Feature activation+0.498
wasn
Token wasn
Feature activation+1.629
't
Token't
Feature activation+2.038
creative
Token creative
Feature activation+0.371
enough
Token enough
Feature activation+0.437
for
Token for
Feature activation+0.554
me
Token me
Feature activation+0.000
.
Token.
Feature activation+0.000

INTERVAL 1.749 - 1.999
CONTAINS 0.001%

We
Token We
Feature activation+0.000
believe
Token believe
Feature activation+0.000
that
Token that
Feature activation+0.437
it
Token it
Feature activation+1.468
is
Token is
Feature activation+1.972
an
Token an
Feature activation+1.823
ideal
Token ideal
Feature activation+0.558
place
Token place
Feature activation+1.074
to
Token to
Feature activation+1.066
show
Token show
Feature activation+0.298
the
Token the
Feature activation+0.000
not
Token not
Feature activation+0.000
believe
Token believe
Feature activation+0.000
that
Token that
Feature activation+0.028
the
Token the
Feature activation+0.323
coin
Token coin
Feature activation+0.222
is
Token is
Feature activation+1.893
worth
Token worth
Feature activation+1.296
$
Token $
Feature activation+0.000
1
Token1
Feature activation+0.000
trillion
Token trillion
Feature activation+0.602
and
Token and
Feature activation+0.337
.
Token.
Feature activation+0.000
Bel
Token Bel
Feature activation+0.000
ieving
Tokenieving
Feature activation+0.341
the
Token the
Feature activation+0.527
suspect
Token suspect
Feature activation+1.668
was
Token was
Feature activation+1.775
armed
Token armed
Feature activation+0.183
,
Token,
Feature activation+0.000
"
Token "
Feature activation+0.000
both
Tokenboth
Feature activation+0.000
officers
Token officers
Feature activation+0.000
.
Token.
Feature activation+0.000
We
Token We
Feature activation+0.000
believe
Token believe
Feature activation+0.000
that
Token that
Feature activation+0.437
it
Token it
Feature activation+1.468
is
Token is
Feature activation+1.972
an
Token an
Feature activation+1.823
ideal
Token ideal
Feature activation+0.558
place
Token place
Feature activation+1.074
to
Token to
Feature activation+1.066
show
Token show
Feature activation+0.298
ire
Tokenire
Feature activation+0.000
children
Token children
Feature activation+0.000
who
Token who
Feature activation+0.000
believe
Token believe
Feature activation+0.000
a
Token a
Feature activation+1.096
criminal
Token criminal
Feature activation+1.822
they
Token they
Feature activation+0.501
discover
Token discover
Feature activation+0.000
hiding
Token hiding
Feature activation+0.916
in
Token in
Feature activation+0.000
their
Token their
Feature activation+0.000

INTERVAL 1.500 - 1.749
CONTAINS 0.003%

because
Token because
Feature activation+0.000
Kim
Token Kim
Feature activation+0.000
believes
Token believes
Feature activation+0.000
that
Token that
Feature activation+0.265
will
Token will
Feature activation+1.177
give
Token give
Feature activation+1.521
him
Token him
Feature activation+0.573
his
Token his
Feature activation+0.479
best
Token best
Feature activation+0.266
chance
Token chance
Feature activation+0.027
of
Token of
Feature activation+0.400
regulations
Token regulations
Feature activation+0.000
and
Token and
Feature activation+0.000
continue
Token continue
Feature activation+0.000
to
Token to
Feature activation+0.000
believe
Token believe
Feature activation+0.000
China
Token China
Feature activation+1.537
will
Token will
Feature activation+1.170
be
Token be
Feature activation+1.525
critical
Token critical
Feature activation+0.355
to
Token to
Feature activation+1.179
their
Token their
Feature activation+0.340
continue
Token continue
Feature activation+0.000
to
Token to
Feature activation+0.000
believe
Token believe
Feature activation+0.000
China
Token China
Feature activation+1.537
will
Token will
Feature activation+1.170
be
Token be
Feature activation+1.525
critical
Token critical
Feature activation+0.355
to
Token to
Feature activation+1.179
their
Token their
Feature activation+0.340
growth
Token growth
Feature activation+0.000
and
Token and
Feature activation+0.000
Walker
Token Walker
Feature activation+0.000
felt
Token felt
Feature activation+0.000
that
Token that
Feature activation+0.000
H
Token H
Feature activation+0.000
aged
Tokenaged
Feature activation+0.000
orn
Tokenorn
Feature activation+1.660
was
Token was
Feature activation+1.852
qualified
Token qualified
Feature activation+0.172
for
Token for
Feature activation+0.311
the
Token the
Feature activation+0.000
things
Token things
Feature activation+0.000
discover
Token discover
Feature activation+0.000
hiding
Token hiding
Feature activation+0.916
in
Token in
Feature activation+0.000
their
Token their
Feature activation+0.000
barn
Token barn
Feature activation+0.000
is
Token is
Feature activation+1.651
a
Token a
Feature activation+1.633
reincarn
Token reincarn
Feature activation+0.007
ation
Tokenation
Feature activation+0.624
of
Token of
Feature activation+0.297
Christ
Token Christ
Feature activation+0.000

INTERVAL 1.250 - 1.500
CONTAINS 0.005%

like
Token like
Feature activation+0.000
me
Token me
Feature activation+0.000
I
Token I
Feature activation+0.000
believe
Token believe
Feature activation+0.019
that
Token that
Feature activation+0.429
you
Token you
Feature activation+1.384
care
Token care
Feature activation+0.958
for
Token for
Feature activation+1.189
the
Token the
Feature activation+0.516
full
Token full
Feature activation+0.000
breadth
Token breadth
Feature activation+0.139
believe
Token believe
Feature activation+0.000
special
Token special
Feature activation+0.000
counsel
Token counsel
Feature activation+0.000
Robert
Token Robert
Feature activation+0.000
Mueller
Token Mueller
Feature activation+1.189
is
Token is
Feature activation+1.458
on
Token on
Feature activation+0.252
a
Token a
Feature activation+0.256
witch
Token witch
Feature activation+0.000
hunt
Token hunt
Feature activation+0.080
as
Token as
Feature activation+0.000
the
Token the
Feature activation+0.000
president
Token president
Feature activation+0.000
believes
Token believes
Feature activation+0.000
that
Token that
Feature activation+0.541
we
Token we
Feature activation+1.295
can
Token can
Feature activation+1.312
do
Token do
Feature activation+0.442
some
Token some
Feature activation+0.207
very
Token very
Feature activation+0.000
helpful
Token helpful
Feature activation+0.000
things
Token things
Feature activation+0.000
ins
Tokenins
Feature activation+0.000
.
Token.
Feature activation+0.000
Sell
Token Sell
Feature activation+0.000
ers
Tokeners
Feature activation+0.000
felt
Token felt
Feature activation+0.000
he
Token he
Feature activation+1.307
was
Token was
Feature activation+1.290
already
Token already
Feature activation+0.669
stretched
Token stretched
Feature activation+0.000
with
Token with
Feature activation+0.000
3
Token 3
Feature activation+0.000
center
Token center
Feature activation+0.000
.
Token.
Feature activation+0.000
We
Token We
Feature activation+0.000
believe
Token believe
Feature activation+0.000
that
Token that
Feature activation+0.437
it
Token it
Feature activation+1.468
is
Token is
Feature activation+1.972
an
Token an
Feature activation+1.823
ideal
Token ideal
Feature activation+0.558
place
Token place
Feature activation+1.074
to
Token to
Feature activation+1.066

INTERVAL 1.000 - 1.250
CONTAINS 0.010%

number
Token number
Feature activation+0.000
of
Token of
Feature activation+0.000
voters
Token voters
Feature activation+0.000
who
Token who
Feature activation+0.000
believe
Token believe
Feature activation+0.000
terrorists
Token terrorists
Feature activation+1.176
are
Token are
Feature activation+1.032
winning
Token winning
Feature activation+0.000
the
Token the
Feature activation+0.136
fight
Token fight
Feature activation+0.000
against
Token against
Feature activation+0.000
nor
Token nor
Feature activation+0.000
the
Token the
Feature activation+0.000
Federal
Token Federal
Feature activation+0.000
Reserve
Token Reserve
Feature activation+0.000
believes
Token believes
Feature activation+0.200
that
Token that
Feature activation+1.180
the
Token the
Feature activation+0.564
law
Token law
Feature activation+0.600
can
Token can
Feature activation+1.438
or
Token or
Feature activation+0.143
should
Token should
Feature activation+0.700
is
Token is
Feature activation+0.000
a
Token a
Feature activation+0.000
band
Token band
Feature activation+0.000
who
Token who
Feature activation+0.000
feel
Token feel
Feature activation+0.000
they
Token they
Feature activation+1.106
've
Token've
Feature activation+1.249
earned
Token earned
Feature activation+0.482
the
Token the
Feature activation+0.000
right
Token right
Feature activation+0.000
to
Token to
Feature activation+0.000
team
Token team
Feature activation+0.000
lost
Token lost
Feature activation+0.000
but
Token but
Feature activation+0.000
thought
Token thought
Feature activation+0.092
it
Token it
Feature activation+0.948
was
Token was
Feature activation+1.150
neat
Token neat
Feature activation+0.575
that
Token that
Feature activation+0.000
his
Token his
Feature activation+0.000
team
Token team
Feature activation+0.000
from
Token from
Feature activation+0.000
Griffin
Token Griffin
Feature activation+0.000
thought
Token thought
Feature activation+0.000