TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.462

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

<|endoftext|> perspective . " What is true , though ,

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

. " What is true , though , is that whatever

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.08

Head 1: 0.09

Head 2: 0.08

Head 3: 0.08

Head 4: 0.09

Head 5: 0.08

Head 6: 0.09

Head 7: 0.09

Head 8: 0.08

Head 9: 0.07

Head 10: 0.08

Head 11: 0.08

Positive logits

Archae1.71

Topics1.56

Tradition1.56

‎1.54

Scholars1.53

lishes1.51

Students1.48

Abbey1.48

Literary1.48

Lect1.47

Historic1.47

Arist1.47

Fellow1.46

Faculty1.45

Father1.44

Founding1.42

Dram1.40

Philos1.40

Students1.40

Extension1.38

Negative logits

amba-1.69

underest-1.57

thrott-1.56

epad-1.52

casinos-1.51

apo-1.50

userc-1.47

atel-1.46

puff-1.44

tricked-1.43

overpowered-1.42

ikuman-1.42

inclined-1.41

asia-1.39

asin-1.38

hooked-1.38

��-1.36

thirst-1.35

incurred-1.35

skew-1.34

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

to gain advantage . CH E AT to Improve your hand

unt men ] train for months to make sure that every

factor Sunday , the Jaguars were in it . When he

and potential fines stemming from the 2010 disaster , emergency responders

presumed he was an atheist , and the apparent intent was

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 1: Uninterpretable

TOP ACTIVATIONS
MAX = 5.379

. Ċ Ċ S omew here between 10 and 40 of

. Ċ Ċ S omew here into their swim , the

. âĢ ľ S omew here if you look along most

? Ċ Ċ S omew here between an hour and twenty

. Ċ Ċ S omew here out there âĢ¦. Ċ Ċ

. Ċ Ċ S omew hat ironically , separate studies have

. Ċ Ċ S omew hat inevitably , the bridge takes

design Ċ Ċ s omew here on our network drives that

Ċ [ S ] omew here deep inside her she knew

They 're " s omew hat unw ield y ," said

. Ċ Ċ S omew hat good news so for England

term decision " s omew here deep in government " to

first six ( s omew hat controversial ) years , the

Ļ s ( s omew hat over designed ) machines from

. Ċ Ċ S omew here during the next decade good

and budget --- s omew here affordable that 's warmer (

like a ( s omew hat ) sensible use of time

is âĢ ľ s omew hat misleading in that it implies

is âĢ ľ s omew hat important , âĢ Ŀ to

have a ( s omew hat tarn ished ) set of

Top DFA by src position
MAX = 6.777

too . Ċ Ċ S omew here between 10 and 40

report . Ċ Ċ S omew here into their swim ,

said . âĢ ľ S omew here if you look along

next ? Ċ Ċ S omew here between an hour and

other . Ċ Ċ S omew here out there âĢ¦. Ċ

assumption . Ċ Ċ S omew hat ironically , separate studies

miles . Ċ Ċ S omew hat inevitably , the bridge

or design Ċ Ċ s omew here on our network drives

Ċ Ċ [ S ] omew here deep inside her she

. They 're " s omew hat unw ield y ,"

team . Ċ Ċ S omew hat good news so for

- term decision " s omew here deep in government "

the first six ( s omew hat controversial ) years ,

âĢ Ļ s ( s omew hat over designed ) machines

C . Ċ Ċ S omew here during the next decade

ings and budget --- s omew here affordable that 's warmer

seem like a ( s omew hat ) sensible use of

line is âĢ ľ s omew hat misleading in that it

it is âĢ ľ s omew hat important , âĢ Ŀ

we have a ( s omew hat tarn ished ) set

Decoder Weights Distribution

Head 0: 0.29

Head 1: 0.05

Head 2: 0.03

Head 3: 0.02

Head 4: 0.03

Head 5: 0.03

Head 6: 0.32

Head 7: 0.03

Head 8: 0.03

Head 9: 0.03

Head 10: 0.03

Head 11: 0.10

Positive logits

stranger1.78

ebin1.65

mum1.62

FTWARE1.60

neighbor1.59

begging1.58

AU1.57

ILCS1.51

kindred1.51

neighbour1.51

strangers1.49

diamonds1.47

ransom1.45

fleeing1.45

loo1.44

gunshots1.44

instability1.43

WARE1.42

pause1.41

ersen1.39

Negative logits

omics-1.72

mone-1.63

Anthem-1.61

otation-1.60

uph-1.58

isd-1.56

anas-1.55

ALD-1.53

�-1.52

otom-1.52

surg-1.51

brand-1.51

onis-1.50

ラン-1.50

pex-1.49

phas-1.49

フォ-1.47

ises-1.47

ald-1.46

atro-1.43

INTERVAL 4.841 - 5.379
CONTAINS 0.000%

. Ċ Ċ S omew here out there âĢ¦. Ċ Ċ

. Ċ Ċ S omew hat ironically , separate studies have

. Ċ Ċ S omew here between 10 and 40 of

. âĢ ľ S omew here if you look along most

? Ċ Ċ S omew here between an hour and twenty

INTERVAL 4.304 - 4.841
CONTAINS 0.000%

. Ċ Ċ S omew hat inevitably , the bridge takes

Ċ [ S ] omew here deep inside her she knew

design Ċ Ċ s omew here on our network drives that

. Ċ Ċ S omew hat good news so for England

They 're " s omew hat unw ield y ," said

INTERVAL 3.766 - 4.304
CONTAINS 0.000%

Ļ s ( s omew hat over designed ) machines from

like a ( s omew hat ) sensible use of time

is âĢ ľ s omew hat important , âĢ Ŀ to

and budget --- s omew here affordable that 's warmer (

have a ( s omew hat tarn ished ) set of

INTERVAL 3.228 - 3.766
CONTAINS 0.000%

open classes ( s omew hat less - gr aciously known

" or " s omew hat " worried about Trump having

responded âĢ ľ s omew hat unfair âĢ Ŀ or âĢ

] Ċ Ċ S omew here between a few days to

Virgin Media ( s omew hat ironically ) has been hook

INTERVAL 2.690 - 3.228
CONTAINS 0.000%

are just ( s omew hat lump y ) gravy .

INTERVAL 2.152 - 2.690
CONTAINS 0.000%

disk management tool . Somew hat more versatile are Menu et

as their guardian . Somew hat ironically , however , while

Im press ions âĢĵ Somew hat of a one dimensional F

, âĢ ľ s omew hat appealing âĢ Ŀ , âĢ

INTERVAL 1.614 - 2.152
CONTAINS 0.000%

Ċ Ċ S omew hat ironically , separate studies have found

s omew hat unw ield y ," said the executive director

to low body . Somew hat ast ring ent . Ċ

INTERVAL 1.076 - 1.614
CONTAINS 0.000%

evidence for inflation is circumst ant ial , which is why

a better player . Stead y improvements in every part of

past was based on circumst ant ial evidence because there was

CAN TE LL SOM ETH ING IS OFF . Know yourself

accumulating a mountain of circumst ant ial evidence that the pers

INTERVAL 0.538 - 1.076
CONTAINS 0.000%

ictional Games âĢ Ļ SOM A , whose creators also cite

acceptable speech have grown indist inct . We reap the whirlwind

much for the parts SOM A cannibal ized . How does

and pur ification . Somewhere along the way , the tradition

Media ( s omew hat ironically ) has been hook ing

INTERVAL 0.000 - 0.538
CONTAINS 99.999%

in a good light . Two of them show Mer ck

ledge $ 15 or more and you will get the previous

use . Ċ Ċ The ruling by the Sacramento - based

. âĢ ľ I have no more memories than from the

amounts of it consisting of min enges cho ÃŁ Ċ Ċ

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 2: In word / title starting with " Som"

TOP ACTIVATIONS
MAX = 3.276

their money from wealthy Som alis abroad , as well as

attraction the the Royal Som mer hus meet and greet area

elements ( except the som a ) with shared electronic circuits

to any visual , som atos ens ory , or auditory

community right here in Som erville . Ċ Ċ We also

wrist band designed by Som atic Labs relies instead on touch

area of the Royal Som mer hus we see that the

will inger and Jacob Som mer ) who represented Yahoo in

inter play . The som atos ens ory cortex ( the

team âĢ Ļ s som n ol ent finish after winning

see you at the Som erville Armory on March 2 nd

stir from its apparent som n ol ence to send this

arrested on 93 in Som erville # 7 news pic .

of the room from som ber to tear ful . Ċ

1992 killings of a Som erville family . His co -

capt ivated audiences by som ers ault ing between five different

ainer thanks Re ena Som mer ; Jeffrey Witt man ,

British sector of the Som me , she set out by

any visual , som atos ens ory , or auditory stimulation

the trenches at the Som me in Picard ie , Northern

Top DFA by src position
MAX = 5.412

of their money from wealthy Som alis abroad , as well

After attraction the the Royal Som mer hus meet and greet

neural elements ( except the som a ) with shared electronic

respond to any visual , som atos ens ory , or

learning community right here in Som erville . Ċ Ċ We

new wrist band designed by Som atic Labs relies instead on

greet area of the Royal Som mer hus we see that

Z will inger and Jacob Som mer ) who represented Yahoo

this inter play . The som atos ens ory cortex (

the team âĢ Ļ s som n ol ent finish after

to see you at the Som erville Armory on March 2

to stir from its apparent som n ol ence to send

ors arrested on 93 in Som erville # 7 news pic

mood of the room from som ber to tear ful .

<|endoftext|> 1992 killings of a Som erville family . His co

he capt ivated audiences by som ers ault ing between five

Expl ainer thanks Re ena Som mer ; Jeffrey Witt man

the British sector of the Som me , she set out

respond to any visual , som atos ens ory , or

to the trenches at the Som me in Picard ie ,

Decoder Weights Distribution

Head 0: 0.38

Head 1: 0.05

Head 2: 0.04

Head 3: 0.03

Head 4: 0.05

Head 5: 0.06

Head 6: 0.10

Head 7: 0.05

Head 8: 0.03

Head 9: 0.03

Head 10: 0.04

Head 11: 0.14

Positive logits

aults1.70

owler1.67

lav1.54

acht1.46

yout1.43

oult1.41

otor1.41

illi1.40

awk1.40

olor1.39

oké1.37

onica1.37

illions1.36

osity1.36

pel1.36

lear1.35

ault1.34

brink1.34

gee1.33

uple1.33

Negative logits

��-1.54

��-1.46

Reincarn-1.36

fructose-1.36

��-1.34

Pigs-1.34

Mankind-1.31

Jordanian-1.29

Vanderbilt-1.29

ICLE-1.28

Personnel-1.28

CTV-1.27

IDA-1.26

Winc-1.26

�-1.25

Gallup-1.25

Skydragon-1.24

Chrys-1.23

BLIC-1.23

quished-1.21

INTERVAL 2.949 - 3.276
CONTAINS 0.000%

attraction the the Royal Som mer hus meet and greet area

will inger and Jacob Som mer ) who represented Yahoo in

to any visual , som atos ens ory , or auditory

elements ( except the som a ) with shared electronic circuits

wrist band designed by Som atic Labs relies instead on touch

INTERVAL 2.621 - 2.949
CONTAINS 0.000%

arrested on 93 in Som erville # 7 news pic .

inter play . The som atos ens ory cortex ( the

see you at the Som erville Armory on March 2 nd

of the room from som ber to tear ful . Ċ

team âĢ Ļ s som n ol ent finish after winning

INTERVAL 2.293 - 2.621
CONTAINS 0.000%

, Mu ir and Som erville have all said that evidence

Ċ Ċ This seemingly som n amb ul istic behavioral pattern

1992 killings of a Som erville family . His co -

feature in the right som atos ens ory cortex , which

White House . Chip Som ode v illa / Getty Images

INTERVAL 1.966 - 2.293
CONTAINS 0.000%

physical activities such as som ers aults , cart whe els

. Consider this : SOM EW H ERE in our dense

level in the som atos ens ory cortex probably contributes to

âĢ Ļ s som n ol ent finish after winning two

Ċ Ċ Contact Heather Som erville at 510 - 208 -

INTERVAL 1.638 - 1.966
CONTAINS 0.000%

free educational opportunities in Som erville in the future , such

pretty hilarious one for Som vil ay , with a lot

the mood was more som ber . Ċ Ċ On the

the the Royal Som mer hus meet and greet area .

from a daemon . Som n ium presents a detailed imaginative

INTERVAL 1.311 - 1.638
CONTAINS 0.000%

herb With ania som n ifer a ( W SE )

in the right som atos ens ory cortex , which receives

visual , som atos ens ory , or auditory stimulation indicative

Consider this : SOM EW H ERE in our dense universe

im n Ã¤ ch sten Som mer ge ge ben en

INTERVAL 0.983 - 1.311
CONTAINS 0.000%

ch sten Som mer ge ge ben en falls f Ã¼r

Ċ S omed ay , Som eday , Maybe is a

ONE . IS THAT SOM ETH ING YOU ARE PLAN NING

end to <|endoftext|> About Som eday , Som eday , Maybe

With ania som n ifer a ( W SE ) in

INTERVAL 0.655 - 0.983
CONTAINS 0.000%

sam ma s ak som att an pass a sig social

<|endoftext|> 1992 killings of a Som erville family . His co

âĢ Ŀ and that SOC OM doesn âĢ Ļ t âĢ

of their money from wealthy Som alis abroad , as well

Ċ Ċ Only at Som nia Main Store Ċ Ċ W

INTERVAL 0.328 - 0.655
CONTAINS 0.000%

[ Photo by Chris Sew ell ] <|endoftext|> Poll Ċ Ċ

ab as har som sy f te att sam la sam

, Som eday , Maybe Ċ Ċ NEW YORK TIM ES

Ã ® n se amn Ä ĥ c Ä ĥ tre

Ã¥ n Ã¥ got som man e gent lig en int

INTERVAL 0.000 - 0.328
CONTAINS 99.998%

so make sure to snap your photos and tag @ uber

for violence , but none of the real sense of tragedy

point for Lex âĢ Ļ s technology , though . Consumers

thankful for the city government âĢ Ļ s giveaways this Christmas

up with developers to provide funding to them . Ċ Ċ

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 3: Follows "Def"

TOP ACTIVATIONS
MAX = 4.369

. Ċ Ċ " Def end her from the new invaders

Ċ Dub bed " Def end our Freedom ," organizers say

âĢĶ Digital Ċ Ċ Def end your ancestral castle âĢĶ show

the report entitled " Def ending Disney : Cat aly sts

themes to feature . Def end V ille Two âĢ Ļ

leg . Ċ Ċ Def ending , James Stur man QC

26 52 205 <|endoftext|> Def ending World Champions , European Champions

too . Ċ Ċ Def end : the map must have

OS . Ċ Ċ Def ine ' personal computer ' Ċ

half guard Ċ Ċ Def end against arm lock and Kim

T rap Ċ Ċ Def end Ċ Ċ T oxic Ċ

's L oes ch Def ends Video Of Marines Reported ly

from two games . Def ending champions Bor ussia Dortmund play

Also Read : Trump Def ends Michael Flynn , Bl ames

some cases decades âĢĶ def ended the principles of limited government

Review in iTunes <|endoftext|> Def ending the conglomerate 's stock after

Is it âĢ ľ Def end On Your Own âĢ Ŀ

ST Y LE : Def end deep , fight hard Unlike

. Seek justice . Def end the oppressed . âĢ Ŀ

PRO confirmed . <|endoftext|> Def ending champion Matt Neal is itching

Top DFA by src position
MAX = 5.434

man . Ċ Ċ " Def end her from the new

Ċ Ċ Dub bed " Def end our Freedom ," organizers

4 âĢĶ Digital Ċ Ċ Def end your ancestral castle âĢĶ

in the report entitled " Def ending Disney : Cat aly

other themes to feature . Def end V ille Two âĢ

the leg . Ċ Ċ Def ending , James Stur man

9 26 52 205 <|endoftext|> Def ending World Champions , European

traps too . Ċ Ċ Def end : the map must

latest OS . Ċ Ċ Def ine ' personal computer '

from half guard Ċ Ċ Def end against arm lock and

Ċ T rap Ċ Ċ Def end Ċ Ċ T oxic

CNN 's L oes ch Def ends Video Of Marines Reported

point from two games . Def ending champions Bor ussia Dortmund

Ċ Also Read : Trump Def ends Michael Flynn , Bl

in some cases decades âĢĶ def ended the principles of limited

& Review in iTunes <|endoftext|> Def ending the conglomerate 's stock

Ċ Is it âĢ ľ Def end On Your Own âĢ

Ċ ST Y LE : Def end deep , fight hard

right . Seek justice . Def end the oppressed . âĢ

P PRO confirmed . <|endoftext|> Def ending champion Matt Neal is

Decoder Weights Distribution

Head 0: 0.23

Head 1: 0.05

Head 2: 0.03

Head 3: 0.02

Head 4: 0.04

Head 5: 0.04

Head 6: 0.37

Head 7: 0.02

Head 8: 0.03

Head 9: 0.03

Head 10: 0.03

Head 11: 0.12

Positive logits

DEFENSE1.87

illance1.76

イト1.75

defending1.73

EMENT1.71

POSE1.69

DEF1.65

FLAG1.64

TEAM1.61

FIGHT1.60

ategories1.57

STAT1.55

veland1.55

alion1.53

ATT1.53

Targ1.52

CVE1.51

WARRANT1.50

Defense1.49

Rescue1.49

Negative logits

aji-1.94

orrow-1.72

ophon-1.68

arrang-1.66

ophone-1.66

au-1.66

ither-1.62

ynes-1.62

pots-1.55

ats-1.54

awatts-1.54

Mata-1.52

anian-1.51

olesc-1.50

lio-1.50

oven-1.50

discont-1.50

inge-1.49

onge-1.46

idd-1.45

INTERVAL 3.932 - 4.369
CONTAINS 0.000%

OS . Ċ Ċ Def ine ' personal computer ' Ċ

26 52 205 <|endoftext|> Def ending World Champions , European Champions

. Ċ Ċ " Def end her from the new invaders

too . Ċ Ċ Def end : the map must have

half guard Ċ Ċ Def end against arm lock and Kim

INTERVAL 3.495 - 3.932
CONTAINS 0.000%

local citizens to " def end and protect them , their

Also Read : Trump Def ends Michael Flynn , Bl ames

we have legislation [ def ending the rights of transgender people

ST Y LE : Def end deep , fight hard Unlike

PRO confirmed . <|endoftext|> Def ending champion Matt Neal is itching

INTERVAL 3.058 - 3.495
CONTAINS 0.000%

of what Ċ Ċ def ines a âĢ ľ header âĢ

Ver ifying Double Not Def ined Error = Class . new

's L oes ch Def ends Video Of Marines Apparently Ur

Ċ Ċ All iance Def ending Freedom âĢĶ a legal organization

2015 series standings . Def ending champion Speed âĢ ľ expected

INTERVAL 2.621 - 3.058
CONTAINS 0.000%

Ċ Step 1 . Def ine the business / data class

and addict ing ! Def end your world from the invasion

Ċ How do You Def ine an Animation ? Ċ Ċ

behind such industry - def ining titles as the sci -

hard - to - def end costs . Three examples here

INTERVAL 2.184 - 2.621
CONTAINS 0.000%

of the advocacy group Def ine American applied for protection under

's control Ċ Ċ Def ence : Jake Humph rey took

as a âĢ ľ def ining moment âĢ Ŀ for political

. Some âĢ ľ def enders âĢ Ŀ of the âĢ

plan . Ċ Ċ Def enders of payments say that if

INTERVAL 1.747 - 2.184
CONTAINS 0.000%

acted in self - def ence by killing a drug suspect

Ċ Step 1 . Def ine Hub Tile in X AM

a substantial âĢ ĺ Def ence D ivid end âĢ Ļ

: B Ċ Ċ Def ence : The Jets âĢ Ļ

exec rises Ċ Ċ Def ine a const uctor for cart

INTERVAL 1.311 - 1.747
CONTAINS 0.000%

Ċ Get ready to DEF END THE WORLD ! The biggest

ride and implement the Def ine Test Order ing method .

. continental missile - def ence shield , after famously opting

{ protected override void Def ine Test Order ing () {

during a season - def ining , late - season losing

INTERVAL 0.874 - 1.311
CONTAINS 0.000%

0 ". Ċ Ċ Def ensive End Ċ Ċ I don

BC values from . Def aults are / proc / bc

Ċ The now - def unct online drug market Silk Road

shooting . Ċ Ċ Correct ed troop description of Veteran Late

for a now - def unct travel network , the third

INTERVAL 0.437 - 0.874
CONTAINS 0.001%

Chargers . Ċ Ċ DEF ENSE Ċ Ċ Joe y B

light : Ċ Ċ Def iant 700 l umen focusing LED

the Texas Longh orns defends against the North Dakota State B

, Kelly Campbell ( def initely a human ). Ċ Ċ

. Ċ Ċ " Correct ions has a duty of care

INTERVAL 0.000 - 0.437
CONTAINS 99.998%

as true now as it has been for some twenty years

party insiders and elected officials , had four gubernatorial contenders to

, but you could definitely see them as president . Hillary

as Joseph Mann , is described as a peaceful young man

invoke Him , my court open its sessions with ` God

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 4: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.334

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

<|endoftext|> perspective . " What is

<|endoftext|> perspective . " What is true , though , is

<|endoftext|> perspective . " What is

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

<|endoftext|> perspective . " What is

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.10

Head 1: 0.09

Head 2: 0.07

Head 3: 0.06

Head 4: 0.09

Head 5: 0.09

Head 6: 0.08

Head 7: 0.09

Head 8: 0.07

Head 9: 0.07

Head 10: 0.07

Head 11: 0.12

Positive logits

Illuminati1.79

RAW1.73

Sage1.51

Mayer1.43

Bitcoin1.40

Sorcerer1.40

Paste1.39

Nass1.37

xus1.37

enium1.36

Enlightenment1.35

Magical1.35

Milo1.35

Domain1.35

Ascension1.34

Amelia1.33

Psy1.32

hid1.32

Canaver1.32

Newman1.32

Negative logits

nikov-1.82

etheless-1.76

dra-1.72

theless-1.47

foes-1.42

++;-1.40

veto-1.40

inel-1.39

appease-1.39

ops-1.38

poon-1.38

�-1.37

retion-1.36

challengers-1.36

ivably-1.34

starve-1.33

IELD-1.33

zeb-1.33

fps-1.32

blight-1.31

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

going to change our landscape . âĢ Ŀ Ċ Ċ Ms

us love z a n ju ", nav odi Ver an

a main theater for Egyptian - Israeli fighting over the course

program ; they use beans from Heart in Portland . I

believes it âĢ Ļ s better for business to own more

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 5: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.225

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

that whatever policy prescriptions that we 've been proposing don 't

<|endoftext|> perspective . " What is

<|endoftext|> perspective . " What is true ,

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

<|endoftext|> perspective . " What is

, though , is that whatever policy prescriptions that we 've

<|endoftext|> perspective . " What is

Decoder Weights Distribution

Head 0: 0.08

Head 1: 0.10

Head 2: 0.07

Head 3: 0.08

Head 4: 0.08

Head 5: 0.10

Head 6: 0.09

Head 7: 0.07

Head 8: 0.09

Head 9: 0.08

Head 10: 0.07

Head 11: 0.09

Positive logits

iar1.67

Balk1.55

Levant1.52

aid1.50

icit1.49

Flan1.49

Í1.49

cit1.48

orer1.48

predec1.47

Downing1.46

uers1.46

maritime1.46

purch1.44

aria1.44

WARE1.42

ici1.42

colle1.41

nell1.40

cryptoc1.38

Negative logits

gradient-1.78

puff-1.70

70710-1.67

heats-1.56

heater-1.55

therm-1.51

goggles-1.51

transformer-1.50

mouse-1.50

glove-1.49

temperature-1.49

magnet-1.48

experiments-1.46

experiment-1.46

holster-1.46

rats-1.46

oxy-1.44

hood-1.44

mirror-1.44

radiator-1.43

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

of Cre recomb in ase ( T yr - Cre (

D IE ! âĢ Ŀ Ċ Ċ Picture by Matt Gill

crisis would have ended if Chinese capital flows had been as

't want to give it away ," Trump claimed . Ċ

also be a candidate for individual ized treatment . <|endoftext|> UPDATE

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 6: " human" token

TOP ACTIVATIONS
MAX = 1.749

climate change st alks the human future âĢĵ and it is

warming : Ċ Ċ The human fingerprint in atmospheric carbon dioxide

here boils down to whether human ism is essentially atheist .

SU IT : Making the human - sized suits are not

Although tropical deforestation meets some human needs , it also has

the natural lens of the human eye . Ċ Ċ The

ism , I mentioned that human ism doesn âĢ Ļ t

mandatory âĢ ĺ water for human consumption âĢ Ļ specified for

of the scientific evidence for human - induced climate change ,

, concluding that " the human - memory - based method

what does this mean for human health ?, in light of

a little thicker than a human hair âĢĶ whose brightness can

person . Ċ Ċ Every human has a different headphone âĢ

the human experience . The human experience is 360 degrees ,

the very circumstances that dominated human evolution . It was this

of scientific facts about the human body including a discussion of

insign ific ance of individual human life in the grand scheme

views is destructive to the human experience . Religion gives answers

eb a - never a human person . It is just

for the advancement of the human domain . Ċ Ċ Here

Top DFA by src position
MAX = 3.567

climate change st alks the human future âĢĵ and it is

warming : Ċ Ċ The human fingerprint in atmospheric carbon dioxide

here boils down to whether human ism is essentially atheist .

SU IT : Making the human - sized suits are not

Although tropical deforestation meets some human needs , it also has

the natural lens of the human eye . Ċ Ċ The

ism , I mentioned that human ism doesn âĢ Ļ t

mandatory âĢ ĺ water for human consumption âĢ Ļ specified for

of the scientific evidence for human - induced climate change ,

, concluding that " the human - memory - based method

what does this mean for human health ?, in light of

a little thicker than a human hair âĢĶ whose brightness can

person . Ċ Ċ Every human has a different headphone âĢ

Ċ This is not the human experience . The human experience

the very circumstances that dominated human evolution . It was this

of scientific facts about the human body including a discussion of

insign ific ance of individual human life in the grand scheme

views is destructive to the human experience . Religion gives answers

eb a - never a human person . It is just

for the advancement of the human domain . Ċ Ċ Here

Decoder Weights Distribution

Head 0: 0.17

Head 1: 0.05

Head 2: 0.03

Head 3: 0.02

Head 4: 0.06

Head 5: 0.13

Head 6: 0.11

Head 7: 0.09

Head 8: 0.03

Head 9: 0.02

Head 10: 0.03

Head 11: 0.26

Positive logits

livion1.22

Prix1.21

Queensland1.17

ised1.15

Bengal1.14

ESE1.14

Chennai1.13

stros1.12

Bombay1.11

frog1.11

esc1.11

Aberdeen1.10

Voices1.09

dogs1.09

readable1.08

Rugby1.07

ageddon1.06

gow1.05

monkey1.04

Kerala1.04

Negative logits

hedon-1.15

plet-1.11

energ-1.10

nonviolent-1.07

aminer-1.05

electrom-1.04

coefficients-1.04

outgoing-1.02

stabilize-1.02

insure-1.01

diplom-1.01

ricular-1.01

uster-1.01

oft-1.01

iman-1.00

leng-1.00

Mach-1.00

moderately-1.00

shorten-1.00

canceled-0.99

INTERVAL 1.574 - 1.749
CONTAINS 0.000%

climate change st alks the human future âĢĵ and it is

warming : Ċ Ċ The human fingerprint in atmospheric carbon dioxide

INTERVAL 1.399 - 1.574
CONTAINS 0.000%

mandatory âĢ ĺ water for human consumption âĢ Ļ specified for

here boils down to whether human ism is essentially atheist .

a perfect metaphor for the human cycle of life . Ċ

possible gift âĢĵ a more human face . âĢ Ŀ Ċ

, concluding that " the human - memory - based method

INTERVAL 1.224 - 1.399
CONTAINS 0.002%

- thirds of the total human population by 2050 , experts

everything in common with a human z yg ote except that

Ŀ In my definition of human ism , I mentioned that

now three states for a human emoji : Ċ Ċ Base

âĢ ľ Everybody deserves basic human respect and civil rights and

INTERVAL 1.049 - 1.224
CONTAINS 0.005%

, we learned about the human horrors of the war ,

of courts hip , a human courts hip display or a

of that mission . Every human being in the world went

the civil , if not human , rights of state workers

it means to be a human being , in the kinds

INTERVAL 0.874 - 1.049
CONTAINS 0.006%

we would never defend its human rights based solely on that

is a downside to every human trait . Let âĢ Ļ

a wound would make the human response more like a sal

to the public about how human - induced climate change has

able to grow an entire human body in the lab ,

INTERVAL 0.700 - 0.874
CONTAINS 0.004%

and " it is a human impulse to want to protect

Ċ Ċ Ever since the Human Gen ome Project , which

ite Jem ima Khan and human rights campaigner Bian ca J

." The trouble with fake human beings is real human beings

than hand held weapons and human bodies . Ċ Ċ Su

INTERVAL 0.525 - 0.700
CONTAINS 0.003%

Special Rapp ort eur on Human Rights on Monday that the

qualifies for financial assistance , Human itarian Aid Commissioner Krist al

research opportunities for veterinary and human medical uses . âĢ Ŀ

by deliberately using civilians as human shields . ( al j

act in is approved for human use in a few years

INTERVAL 0.350 - 0.525
CONTAINS 0.002%

of the High Commissioner for Human Rights . Ċ Ċ Since

of its appearance . A human - like appearance gives a

. S . Health and Human Services Secretary Kathleen Se bel

significantly different from the UN Human Rights Council âĢ Ļ s

the Department of Health and Human Services from tapping into other

INTERVAL 0.175 - 0.350
CONTAINS 0.002%

mammals or any sentient non human is morally objectionable . But

in the Universal Declaration of Human Rights . Ċ Ċ Based

acion es prev ias por hom ic id ios d ol

draw enormous attention to the humanitarian crisis that was engulf ing

Darwin had made post - human ism possible : first ,

INTERVAL 0.000 - 0.175
CONTAINS 99.977%

for the government but apparent âĢ ĺ l ack of skills

bigotry in any form , âĢ Ŀ despite the condemnation of

- hand lane unless they are overt aking , turning right

Ļ t have what Kenny has : the brass band backing

i 2 p update _ 0 . 8 . zip b

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 7: " Jackson" token

TOP ACTIVATIONS
MAX = 1.372

time . Ċ Ċ Chief Jackson âĢ Ļ s press statement

Ċ Sam uel L . Jackson 's Nick Fury and Guardians

morning . Ċ Ċ The Jackson County Sheriff 's Office says

laid a foundation . Hue Jackson said that this 1 -

aj ara . Ċ Ċ Jackson appears to have been living

. Ċ Ċ H ue Jackson is no stranger to working

in custody Tuesday at the Jackson County jail . A phone

[ 37 ] Dr . Jackson was appointed Phys ician to

him ," said eyewitness Larry Jackson . Y ol anda Young

procedure . Ċ Ċ A Jackson County Sheriff 's Office press

Ċ Ċ L ion el Jackson shows of some of his

Ċ Ċ 23 46 Survivor Jackson tells the BBC he has

of taxpayers continues : Jackson ville just agreed to pay $

Ŀ Ċ Ċ But while Jackson âĢ Ļ s high -

Ŀ Ċ Ċ Meanwhile , Jackson also went on to briefly

to the draft , Phil Jackson 's insistence Jo ak im

Small as it is , Jackson really is big next to

field ." <|endoftext|> Luc as Jackson / Reuters A woman and

The question now is whether Jackson will see enough from Griffin

NBA Draft . Ċ Ċ Jackson , an intense competitor and

Top DFA by src position
MAX = 2.696

of Michael Brown , which Jackson refused to release at that

Ċ Sam uel L . Jackson 's Nick Fury and Guardians

morning . Ċ Ċ The Jackson County Sheriff 's Office says

laid a foundation . Hue Jackson said that this 1 -

aj ara . Ċ Ċ Jackson appears to have been living

. Ċ Ċ H ue Jackson is no stranger to working

in custody Tuesday at the Jackson County jail . A phone

[ 37 ] Dr . Jackson was appointed Phys ician to

him ," said eyewitness Larry Jackson . Y ol anda Young

procedure . Ċ Ċ A Jackson County Sheriff 's Office press

Ċ Ċ L ion el Jackson shows of some of his

Ċ Ċ 23 46 Survivor Jackson tells the BBC he has

ing of taxpayers continues : Jackson ville just agreed to pay

Ŀ Ċ Ċ But while Jackson âĢ Ļ s high -

Ŀ Ċ Ċ Meanwhile , Jackson also went on to briefly

to the draft , Phil Jackson 's insistence Jo ak im

Small as it is , Jackson really is big next to

field ." <|endoftext|> Luc as Jackson / Reuters A woman and

The question now is whether Jackson will see enough from Griffin

NBA Draft . Ċ Ċ Jackson , an intense competitor and

Decoder Weights Distribution

Head 0: 0.16

Head 1: 0.04

Head 2: 0.05

Head 3: 0.03

Head 4: 0.06

Head 5: 0.09

Head 6: 0.16

Head 7: 0.07

Head 8: 0.04

Head 9: 0.03

Head 10: 0.05

Head 11: 0.24

Positive logits

*/1.30

stead1.30

iffin1.29

)=(1.24

aide1.23

Lumpur1.23

anship1.21

compet1.20

Correct1.19

Mata1.19

counseling1.15

preferably1.13

Username1.11

Sterling1.11

lynn1.10

soDeliveryDate1.09

ijah1.08

aeper1.07

.""1.07

Seym1.06

Negative logits

rums-1.25

dropping-1.23

toggle-1.23

leaps-1.19

ties-1.19

ドラゴン-1.17

gobl-1.16

latable-1.14

Breach-1.14

idges-1.14

vre-1.13

WT-1.12

usions-1.11

clusions-1.11

ⓘ-1.11

avorite-1.11

ciation-1.11

quila-1.10

collect-1.09

paranormal-1.08

INTERVAL 1.235 - 1.372
CONTAINS 0.000%

time . Ċ Ċ Chief Jackson âĢ Ļ s press statement

INTERVAL 1.097 - 1.235
CONTAINS 0.000%

morning . Ċ Ċ The Jackson County Sheriff 's Office says

Ċ Sam uel L . Jackson 's Nick Fury and Guardians

laid a foundation . Hue Jackson said that this 1 -

aj ara . Ċ Ċ Jackson appears to have been living

INTERVAL 0.960 - 1.097
CONTAINS 0.000%

. Ċ Ċ H ue Jackson is no stranger to working

[ 37 ] Dr . Jackson was appointed Phys ician to

in custody Tuesday at the Jackson County jail . A phone

of taxpayers continues : Jackson ville just agreed to pay $

Ċ Ċ L ion el Jackson shows of some of his

INTERVAL 0.823 - 0.960
CONTAINS 0.000%

ľ Mind games : Phil Jackson âĢ Ļ s Long Strange

Ċ " This season ," Jackson said , " is over

the game ." Ċ Ċ Jackson 's comments may have been

, myself , only know Jackson because , for two -

17 - month - old Jackson . Ċ Ċ Bec oming

INTERVAL 0.686 - 0.823
CONTAINS 0.000%

about that . I trust Jackson 's work with quarterbacks ,

weight . Ċ Ċ Jackson ville State , ranked No .

G ee , if Michael Jackson can terror ize his girlfriend

also have to understand that Jackson is trying to mold Griffin

s readers . Ċ Ċ Jackson <|endoftext|> I took a flyer

INTERVAL 0.549 - 0.686
CONTAINS 0.001%

: DE Dem et rius Jackson of Miami Booker T .

War Department with some of Jackson âĢ Ļ s revelations .

at W L BT of Jackson , Miss . Ċ Ċ

Ċ Ċ Court documents show Jackson was communicating with a third

<|endoftext|> stricter drug policy . Jackson , who will not return

INTERVAL 0.412 - 0.549
CONTAINS 0.001%

a private jet and Michael Jackson memor abilia . Ċ Ċ

station W L BT of Jackson that they got the news

with John Walsh " featured Jackson 's case in July .

1 corners who visited . Jackson at No . 13 is

. If Tra ev on Jackson returns as expected , it

INTERVAL 0.274 - 0.412
CONTAINS 0.001%

the New Horizon Church in Jackson is affiliated with a super

the crime . Ċ Ċ Jackson was granted bail until his

<|endoftext|> for D ese an Jackson . Ċ Ċ *** Ċ

Welcome to âĢ ľ Josh Jackson Decision 2016 . âĢ Ŀ

Ļ s MP , Robert Jackson , who has since retired

INTERVAL 0.137 - 0.274
CONTAINS 0.000%

, Triumph covering the Michael Jackson trial , Triumph meeting Chris

theft , police have accused Jackson of obtaining the monkey from

<|endoftext|> ) and Nicholas Austin Jackson shout at St . Louis

Ċ Ċ 4 th A Jackson singled to center , V

memorable rant at McN ally Jackson Book store , in S

INTERVAL 0.000 - 0.137
CONTAINS 99.997%

News while it 's up to a gun show owner to

have been ads far worse . Ċ Ċ Here are 10

IDE , and have a ton more planned over the next

Load ( 0 ) -- compute diffusion local e bb update

get an email from my school telling me I need to

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 8: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.638

been proposing don 't reach , are not heard by ,

<|endoftext|> perspective . " What is

Decoder Weights Distribution

Head 0: 0.15

Head 1: 0.15

Head 2: 0.09

Head 3: 0.06

Head 4: 0.08

Head 5: 0.06

Head 6: 0.06

Head 7: 0.07

Head 8: 0.05

Head 9: 0.07

Head 10: 0.07

Head 11: 0.09

Positive logits

mortal1.60

fleet1.55

mission1.51

runtime1.50

rede1.47

whit1.44

Cold1.40

ulner1.38

ーク1.36

otten1.35

iors1.34

dwind1.33

certs1.33

woes1.33

tick1.33

redeem1.32

forecast1.31

career1.31

counterparts1.30

env1.28

Negative logits

skirts-1.72

Maw-1.65

raine-1.60

omever-1.60

berra-1.59

Rohing-1.56

Bangl-1.56

Burk-1.56

zag-1.56

anmar-1.54

uador-1.54

Papua-1.53

deen-1.52

Maurit-1.51

uala-1.50

uckland-1.49

Gibraltar-1.49

agall-1.49

vertising-1.48

backer-1.46

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

. <|endoftext|> Big Pharma has rem ade their business models over

, at least âĢĵ when he is being the Spring steen

uns cient ific nature of the test and the small apparent

un interesting delivery will bore your audience or may never reach

to destroy those you love . Ċ Ċ Ap ocalypse the

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 9: tokens related to " penalty” / “punishments”

TOP ACTIVATIONS
MAX = 2.140

in on the fun . Penalty ? No ! <|endoftext|> Sen

that we could have been penal ised for . I 'm

a penalty every time . Penalty ? Yes ! Ċ Ċ

Wednesday , there were severe penalties handed out for the teams

. There were probably seven penalties there and probably more that

We had a lot of penalties . I thought overall it

we bounced back from those penalties . We didn 't let

Ċ Ċ "... Again those penalties just drive you crazy .

book . That 's a penalty every time . Penalty ?

. This time , the penalties didn 't come back to

not saying they weren 't penalties . Over the last two

all - around brawl . Penalty ? No ! Ċ Ċ

the potential for more severe penalties if you don 't file

that 's a penalty . Penalty ? Yes ! Ċ Ċ

legislators said they could add penalties at a later date .

be drive - st opping penalties but our guys stayed focused

job , good effort ! Penalty ? No ! Ċ Ċ

yards . The Jets were penal ized just twice for 9

inside North Korea , as punishment for the rocket launch and

Oregon averaged 8 . 2 penalties per game last season ,

Top DFA by src position
MAX = 2.441

in on the fun . Penalty ? No ! <|endoftext|> Sen

that we could have been penal ised for . I 'm

a penalty every time . Penalty ? Yes ! Ċ Ċ

Wednesday , there were severe penalties handed out for the teams

. There were probably seven penalties there and probably more that

first scrimmage you expect some penalties . We had a lot

we bounced back from those penalties . We didn 't let

Ċ Ċ "... Again those penalties just drive you crazy .

book . That 's a penalty every time . Penalty ?

. This time , the penalties didn 't come back to

, you thought they were penalties ? A : I 'm

all - around brawl . Penalty ? No ! Ċ Ċ

the potential for more severe penalties if you don 't file

that 's a penalty . Penalty ? Yes ! Ċ Ċ

legislators said they could add penalties at a later date .

be drive - st opping penalties but our guys stayed focused

job , good effort ! Penalty ? No ! Ċ Ċ

yards . The Jets were penal ized just twice for 9

inside North Korea , as punishment for the rocket launch and

Oregon averaged 8 . 2 penalties per game last season ,

Decoder Weights Distribution

Head 0: 0.08

Head 1: 0.06

Head 2: 0.04

Head 3: 0.02

Head 4: 0.06

Head 5: 0.18

Head 6: 0.21

Head 7: 0.08

Head 8: 0.03

Head 9: 0.03

Head 10: 0.03

Head 11: 0.18

Positive logits

penalty1.55

punished1.44

penalties1.41

offenses1.35

offence1.35

offences1.31

viol1.30

punish1.29

punishments1.25

whatsoever1.21

aggravated1.21

aggrav1.20

severity1.20

offenders1.19

damages1.18

osity1.17

len1.14

Cruel1.14

ores1.13

punishment1.12

Negative logits

Untitled-1.41

headquartered-1.30

sonian-1.27

atform-1.26

anx-1.22

ACA-1.22

mainline-1.20

aeda-1.18

orpor-1.17

phia-1.16

vironment-1.15

icago-1.14

GREEN-1.13

ufact-1.11

ampunk-1.11

BIL-1.11

iltr-1.10

ebus-1.10

design-1.09

aucuses-1.09

INTERVAL 1.926 - 2.140
CONTAINS 0.000%

in on the fun . Penalty ? No ! <|endoftext|> Sen

INTERVAL 1.712 - 1.926
CONTAINS 0.000%

Wednesday , there were severe penalties handed out for the teams

a penalty every time . Penalty ? Yes ! Ċ Ċ

. There were probably seven penalties there and probably more that

that we could have been penal ised for . I 'm

INTERVAL 1.498 - 1.712
CONTAINS 0.000%

Ċ Ċ "... Again those penalties just drive you crazy .

that 's a penalty . Penalty ? Yes ! Ċ Ċ

book . That 's a penalty every time . Penalty ?

legislators said they could add penalties at a later date .

We had a lot of penalties . I thought overall it

INTERVAL 1.284 - 1.498
CONTAINS 0.000%

punishment and that was the punishment . Ċ Ċ " Unfortunately

of Rice âĢ Ļ s penalties , the video itself is

a flag as well . Penalty ? No ! Ċ Ċ

Ċ Ċ The most severe punishment would be the loss of

quarter . Ċ Ċ Pen alty called on FAM U for

INTERVAL 1.070 - 1.284
CONTAINS 0.001%

severe punishment would be the loss of E akin âĢ Ļ

That wasn 't a forward pass tonight to young Ke le

condemns the use of capital punishment against those with mental and

new provisions would allow for penalties of up to five years

offences punishable with the death penalty . By the beginning of

INTERVAL 0.856 - 1.070
CONTAINS 0.002%

a crime and what its punishment should be . Ċ Ċ

it as some sort of punishment that America has had coming

wing without paying a significant penalty in on - road dri

The penalties are whipping , amput ation of limbs and death

we could have been penal ised for . I 'm not

INTERVAL 0.642 - 0.856
CONTAINS 0.003%

competition watchdog had imposed the penalty on the company . The

costs in pursuing a capital punishment conviction , which usually involves

death penalty in cases of adultery , which disproportionately pun ishes

anded use of the death penalty '. Ċ Ċ Indeed ,

be âĢ ľ severe ly punished . âĢ Ļ âĢ Ļ

INTERVAL 0.428 - 0.642
CONTAINS 0.005%

insisted that only he be punished for his mis d eeds

, though " more severe discipline will be imposed if there

when you implement the death penalty it doesn âĢ Ļ t

the Colts or committing foolish penalties . Ċ Ċ However ,

, less often , to punish those who fail to satisfy

INTERVAL 0.214 - 0.428
CONTAINS 0.009%

presence or use of a weapon , choking , repeated striking

to young Ke le pi scoring under the posts . The

of the society and the sentencing process has to be stern

think twice before repeating the act . So , we decided

, non - violent drug offence [ 18 : 11 mark

INTERVAL 0.000 - 0.214
CONTAINS 99.980%

cause swirling . Not just excessive rubbing , but any contact

2 ] Ċ Ċ However , Rud yard Ki pling was

of friends is building an apparatus with âĢ ľ Ready for

is how we learn new words , our concern is that

accusation against the former FBI director , and this comes mostly

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 10: Uninterpretable

TOP ACTIVATIONS
MAX = 2.633

that âĢ Ļ s ton ally perfect for the characters ,

bit of ton al differ ance between the two speakers ,

prank call , was ton ally off . Guardians of the

spent the extra time ton ing down the document under pressure

website 's tang y ton al mix of a sal acious

ival âĢ Ŀ and ton ally , there are certain similarities

programming were gone or ton ed down drastically . Ċ Ċ

to be done about ton ing down the rhetoric from certain

a competition level yo - yo master ). Ċ Ċ So

killer abs , a ton ed butt and s vel te

Ċ Common - practice ton al theory classical theory was invented

arant ino had already ton ed down the color and spl

the incoming Executive yo - yo Kickstarter that actually fires up

an unders ized yo - yo kick as of late .

Creek 2 is , ton ally speaking , miles away from

the graphics have been ton ed down since the E 3

a good bit of ton al differ ance between the two

that it was a ton ed - down , more realistic

initially sound random and ton al ( and at times faintly

colourful and we actually ton ed it down ," Sud a

Top DFA by src position
MAX = 4.572

move that âĢ Ļ s ton ally perfect for the characters

is a good bit of ton al differ ance between the

a prank call , was ton ally off . Guardians of

âĢĶ spent the extra time ton ing down the document under

the website 's tang y ton al mix of a sal

Surv ival âĢ Ŀ and ton ally , there are certain

night programming were gone or ton ed down drastically . Ċ

needs to be done about ton ing down the rhetoric from

to be a competition level yo - yo master ). Ċ

get killer abs , a ton ed butt and s vel

Ċ Ċ Common - practice ton al theory classical theory was

T arant ino had already ton ed down the color and

really on an unders ized yo - yo kick as of

Wolf Creek 2 is , ton ally speaking , miles away

if the graphics have been ton ed down since the E

is a good bit of ton al differ ance between the

car that it was a ton ed - down , more

can initially sound random and ton al ( and at times

more colourful and we actually ton ed it down ," Sud

Decoder Weights Distribution

Head 0: 0.11

Head 1: 0.04

Head 2: 0.05

Head 3: 0.03

Head 4: 0.05

Head 5: 0.05

Head 6: 0.40

Head 7: 0.05

Head 8: 0.04

Head 9: 0.04

Head 10: 0.04

Head 11: 0.11

Positive logits

utical1.93

artisan1.91

=#1.80

fml1.79

creen1.79

namese1.77

ometric1.77

grading1.76

ebus1.74

sonian1.71

WARE1.70

conserv1.68

NYSE1.65

waivers1.64

realism1.63

framework1.62

bending1.61

achelor1.61

portfolios1.58

ropes1.57

Negative logits

ilers-1.79

idem-1.76

Rally-1.74

Ips-1.72

Elections-1.72

vous-1.71

und-1.69

Mé-1.67

Rousse-1.65

Racing-1.63

Pyr-1.62

poll-1.62

feed-1.61

Dul-1.60

min-1.60

Pont-1.55

NEC-1.53

Rural-1.53

igor-1.52

eming-1.52

INTERVAL 2.369 - 2.633
CONTAINS 0.000%

that âĢ Ļ s ton ally perfect for the characters ,

to be done about ton ing down the rhetoric from certain

spent the extra time ton ing down the document under pressure

programming were gone or ton ed down drastically . Ċ Ċ

bit of ton al differ ance between the two speakers ,

INTERVAL 2.106 - 2.369
CONTAINS 0.000%

arant ino had already ton ed down the color and spl

an unders ized yo - yo kick as of late .

Ċ Common - practice ton al theory classical theory was invented

the incoming Executive yo - yo Kickstarter that actually fires up

a competition level yo - yo master ). Ċ Ċ So

INTERVAL 1.843 - 2.106
CONTAINS 0.000%

a good bit of ton al differ ance between the two

the graphics have been ton ed down since the E 3

might result in the ton al imbalance . Ċ Ċ Different

at the Executive yo - yo Kickstarter . Better yet ,

oman ia , the ton ed - down finale seems to

INTERVAL 1.580 - 1.843
CONTAINS 0.000%

towers , the gorgeous pal aces , Ċ Ċ The solemn

is a regular Target sho pper , says . âĢ ľ

has opted for a ton ed - down aesthetic with the

and MRI scanners to tan ning sal ons and the sale

its ideology , having ton ed down its earlier , od

INTERVAL 1.316 - 1.580
CONTAINS 0.000%

art isans hip and ton ed - down political rhetoric ,

her car while she sho pped at a Walmart in Utah

demographic of a Walmart sho pper is a middle - aged

assault rifles as they sho pped . There is simply no

fastest way to a sho pper 's wallet , and studies

INTERVAL 1.053 - 1.316
CONTAINS 0.000%

business , Al oha Tan ning . Here are a few

âĢ Ŀ said one sho pper . âĢ ľ I think

harmonic direction ality to ton al configurations in which notions like

, and the great pal aces and cities were destroyed or

into miniature yo - y os then you need to get

INTERVAL 0.790 - 1.053
CONTAINS 0.000%

being confronted by a sho pper carrying a concealed weapon .

town to buy a kil o of cocaine from them .

claw ing through the pal aces . The personal assets of

serving as almost a pal let cleans er between each horror

other things that are pric ier on screen than in real

INTERVAL 0.527 - 0.790
CONTAINS 0.000%

to replace the existing kil ogram . But with a few

antenna e are a pal er color .[ 5 ] The

to be live . Ton ally and rhythm ically , it

more about this new miniature yo - yo over in the

Brexit could prove more pal atable than full membership of the

INTERVAL 0.263 - 0.527
CONTAINS 0.000%

their photo at a ke g party , even if they

centers for marriage bro k ering with foreigners exist , most

race with a more pal atable distance that a viewer can

purchased in 100 pound pal lets , so the new color

Investigators also found numerous pal lets of alcohol , including beer

INTERVAL 0.000 - 0.263
CONTAINS 99.998%

ational economy , to grow our population to more than 7

il , s ous un angle d ' une second e

wounds suffered Friday night . Ċ Ċ Er in Meyer details

world it helped create . Ċ Ċ We know that D

from increase the thickness of the handset , which is not

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 11: In text involving money / payments

TOP ACTIVATIONS
MAX = 0.934

price , so you pay per byte , just like you

just like you pay per minute . Same deal . It

, just like you pay per minute . Same deal .

this could mean zero money down deals , and combined with

their potential for profit will increase exponentially should more people sign

, so you pay per byte , just like you pay

new subscriber requires barely any extra effort , their potential for

my initial out lay . Interest rates are low âĢĵ it

100 % on my initial out lay . Interest rates are

deduct ibles ) than a plan covering 70 percent of costs

of 100 % on my initial out lay . Interest rates

covered costs would have much higher cost sharing ( e .

, or you fail to make your lease payment , X

many cases this could mean zero money down deals , and

lay . Interest rates are low âĢĵ it has never been

value , I have a return of 100 % on my

subscriber requires barely any extra effort , their potential for profit

would have much higher cost sharing ( e . g .,

initial out lay . Interest rates are low âĢĵ it has

could mean zero money down deals , and combined with the

Top DFA by src position
MAX = 0.292

price , so you pay per byte , just like you

byte , just like you pay per minute . Same deal

cases this could mean zero money down deals , and combined

their potential for profit will increase exponentially should more people sign

of data for a fixed price , so you pay per

new subscriber requires barely any extra effort , their potential for

my initial out lay . Interest rates are low âĢĵ it

100 % on my initial out lay . Interest rates are

( e . g ., deduct ibles ) than a plan

of 100 % on my initial out lay . Interest rates

current law . A plan paying 58 percent of expected covered

, or you fail to make your lease payment , X

many cases this could mean zero money down deals , and

my initial out lay . Interest rates are low âĢĵ it

value , I have a return of 100 % on my

subscriber requires barely any extra effort , their potential for profit

would have much higher cost sharing ( e . g .,

initial out lay . Interest rates are low âĢĵ it has

could mean zero money down deals , and combined with the

Decoder Weights Distribution

Head 0: 0.05

Head 1: 0.07

Head 2: 0.11

Head 3: 0.04

Head 4: 0.09

Head 5: 0.07

Head 6: 0.12

Head 7: 0.14

Head 8: 0.05

Head 9: 0.05

Head 10: 0.12

Head 11: 0.09

Positive logits

($2.03

bucks1.97

groceries1.94

deductible1.93

coupons1.89

($1.88

insurance1.82

payment1.79

Mortgage1.79

reimbursement1.78

debit1.78

/$1.77

coupon1.74

Insurance1.73

depreciation1.73

mortgage1.72

mortgages1.71

rentals1.70

premium1.70

vouchers1.68

Negative logits

�-1.30

Mosque-1.28

Islam-1.27

militar-1.26

�-1.24

assassinated-1.23

archae-1.21

Cyr-1.21

semin-1.21

ه-1.21

mosques-1.17

Quran-1.15

rahim-1.15

homophobia-1.15

kun-1.15

tragedies-1.14

Arab-1.12

�-1.12

Alam-1.12

Incarn-1.12

INTERVAL 0.841 - 0.934
CONTAINS 0.000%

new subscriber requires barely any extra effort , their potential for

this could mean zero money down deals , and combined with

just like you pay per minute . Same deal . It

my initial out lay . Interest rates are low âĢĵ it

, just like you pay per minute . Same deal .

INTERVAL 0.747 - 0.841
CONTAINS 0.000%

any extra effort , their potential for profit will increase exponentially

more data for the same price as they currently pay .

could mean zero money down deals , and combined with the

Now âĢ Ļ s cheapest plan , and that plan does

. A plan paying 58 percent of expected covered costs would

INTERVAL 0.654 - 0.747
CONTAINS 0.001%

k in value , I have a return of 100 %

hotel stay are along the lines of what is applicable to

you pay per byte , just like you pay per minute

under current law . A plan paying 58 percent of expected

for those who want to try out the technology with fewer

INTERVAL 0.560 - 0.654
CONTAINS 0.002%

" Whether it 's a sur charge or a discount it

app on their phone that shows them when their Power wall

, the $ 35 credit is the price of Dire c

price of Dire c TV Now âĢ Ļ s cheapest plan

use federal money to create high - risk pools as insurers

INTERVAL 0.467 - 0.560
CONTAINS 0.008%

one - way to hand deliver last - minute financial help

000 in benefits around GST credits , Ontario property tax credits

. Those who do not qualify for subsidies may find that

a statement that it was planning to reach out to P

for Sunday Ticket will deter a lot of potential customers .

INTERVAL 0.374 - 0.467
CONTAINS 0.019%

$ 32 , 500 after rebate . Of course , both

queue for a food parcel from the local food bank ,

for one week , I want to see the impact we

$ 5 a month . The energy board told Toronto Hydro

in about three days . Or , if you don 't

INTERVAL 0.280 - 0.374
CONTAINS 0.043%

this will enable you to gain convenient access to withdrawn funds

âĢ Ļ ve ever seen one in action . In our

than rent . Top five markets in which to buy are

incentives . For example , a typical BE V has an

, 000 . His house is " onder water ," as

INTERVAL 0.187 - 0.280
CONTAINS 0.097%

, âĢ ľ setting a target in the 3 - 4

- found ers would be able to live off their basic

their company , they take credit card numbers that people can

with family who are not participating in the tournament we will

off the street . So when Utah tried just handing such

INTERVAL 0.093 - 0.187
CONTAINS 0.207%

5 Gift Card to First 100 Customers ( One to te

it on remotely . As one article suggests : âĢ ľ

âĢ Ļ t need , at least for a while .

to our society of Russian trolling and fake news on Facebook

, 500 bail and released Monday , according to The Smoking

INTERVAL 0.000 - 0.093
CONTAINS 99.623%

meantime , it âĢ Ļ s completing renovations to a space

's eyes gl aze over with this one . I 'm

a practice in how to let go , first of our

defence secretary when the claims were made Tory front ben cher

It will be joined by several low - rise commercial buildings

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 12: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.381

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

we 've been proposing don 't reach , are not heard

<|endoftext|> perspective . " What is

that whatever policy prescriptions that we 've been proposing don 't

<|endoftext|> perspective . " What is

whatever policy prescriptions that we 've been proposing don 't reach

<|endoftext|> perspective . " What is

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

<|endoftext|> perspective . " What is

Decoder Weights Distribution

Head 0: 0.09

Head 1: 0.08

Head 2: 0.09

Head 3: 0.08

Head 4: 0.08

Head 5: 0.10

Head 6: 0.08

Head 7: 0.09

Head 8: 0.08

Head 9: 0.07

Head 10: 0.06

Head 11: 0.09

Positive logits

Levant1.62

romy1.51

Donation1.50

raq1.49

Born1.44

inav1.42

sett1.42

human1.35

Epstein1.33

Hunt1.33

courtroom1.31

chwitz1.31

uthor1.24

abet1.24

isner1.24

telling1.24

Pere1.24

dod1.21

ket1.20

oven1.20

Negative logits

attRot-1.66

eworks-1.65

chuk-1.56

ideo-1.49

ractions-1.47

itself-1.47

CPC-1.44

},{"-1.40

estyles-1.40

independents-1.40

'."-1.40

oya-1.39

steamapps-1.37

ubs-1.37

eatures-1.37

onductor-1.36

ihil-1.36

fps-1.35

AAA-1.35

hiba-1.33

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

years old âĢ¦ stood in front of a wall of Lego

Iraq , he argues that Afghanistan , where the 9 /

rights community has âĢ ľ supported the concept of the Equality

despite the fact that legal construction starts have been prohibited for

7 . Mens Short Ha irst yles + C rop Cut

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 13: Follows “ We”

TOP ACTIVATIONS
MAX = 1.522

Cinem atic Universe . We have designed this Spider man leather

Ċ Ċ âĢ ľ We have people from little tiny kids

and entrepreneurship . " We will innovate a lot in the

. Ċ Ċ " We have <|endoftext|> Enlarge Image Joe Ra

western . âĢ ľ We are tired of the closing of

also added : " We call on all parties to stop

statement . âĢ ľ We will continue our joint efforts to

Ċ Ċ âĢ ľ We have to look at our responsibility

. Ċ Ċ " We have concerns that the law is

Ċ Ċ âĢ ľ We are suffering because there seems to

Ċ Ċ âĢ ľ We have now moved from a growth

by Britain . " We have always said once we have

in 2014 . " We are looking forward to a strong

Tools for Android . We have made several fixes related to

pm AST Ċ Ċ We wish everyone the best of luck

. Ċ Ċ " We believe that as a tactic they

original statement . " We have considered all the evidence carefully

. Ċ Ċ " We are 75 per cent confident that

Ċ Ċ âĢ ľ We have no other platform , even

Edwards . âĢ ľ We are disappointed to lose him .

Top DFA by src position
MAX = 1.765

Marvel Cinem atic Universe . We have designed this Spider man

. Ċ Ċ âĢ ľ We have people from little tiny

startups and entrepreneurship . " We will innovate a lot in

Friday . Ċ Ċ " We have <|endoftext|> Enlarge Image Joe

South western . âĢ ľ We are tired of the closing

He also added : " We call on all parties to

a statement . âĢ ľ We will continue our joint efforts

. Ċ Ċ âĢ ľ We have to look at our

case . Ċ Ċ " We have concerns that the law

. Ċ Ċ âĢ ľ We are suffering because there seems

. Ċ Ċ âĢ ľ We have now moved from a

entirely by Britain . " We have always said once we

sponsor in 2014 . " We are looking forward to a

Dev Tools for Android . We have made several fixes related

59 pm AST Ċ Ċ We wish everyone the best of

disorder . Ċ Ċ " We believe that as a tactic

his original statement . " We have considered all the evidence

ollen . Ċ Ċ " We are 75 per cent confident

. Ċ Ċ âĢ ľ We have no other platform ,

said Edwards . âĢ ľ We are disappointed to lose him

Decoder Weights Distribution

Head 0: 0.44

Head 1: 0.14

Head 2: 0.03

Head 3: 0.02

Head 4: 0.04

Head 5: 0.07

Head 6: 0.03

Head 7: 0.04

Head 8: 0.02

Head 9: 0.03

Head 10: 0.03

Head 11: 0.11

Positive logits

ourselves1.76

aughters1.49

gladly1.48

anwhile1.48

ngth1.47

parted1.45

CONTIN1.44

emonium1.43

Kamp1.40

passionately1.38

terson1.36

nesday1.35

velt1.34

allery1.33

ours1.32

hereby1.30

respectfully1.30

rike1.30

oppers1.30

riad1.29

Negative logits

flu-1.42

ヘ-1.37

bearing-1.30

pex-1.30

stro-1.29

imaginable-1.29

ital-1.29

swing-1.28

esides-1.24

MAL-1.24

license-1.24

otos-1.23

pot-1.21

angular-1.20

job-1.20

ionics-1.20

shape-1.20

startup-1.19

role-1.19

boost-1.18

INTERVAL 1.370 - 1.522
CONTAINS 0.000%

Cinem atic Universe . We have designed this Spider man leather

INTERVAL 1.217 - 1.370
CONTAINS 0.001%

Ċ Ċ âĢ ľ We are against it . There âĢ

to all prisoners . We are very much aware and sympathetic

Ċ Ċ âĢ ľ We have started investigations designed to clarify

of this order . We have filed a plea in the

summer . âĢ ľ We need to put the gear for

INTERVAL 1.065 - 1.217
CONTAINS 0.007%

Rose . âĢ ľ We will continue to work closely with

. Ċ Ċ " We have been overwhelmed with the messages

will create challenges . We are all sad because of it

Ċ Ċ âĢ ľ We are willing to help you with

date : Ċ Ċ We wish you a safe and healthy

INTERVAL 0.913 - 1.065
CONTAINS 0.012%

. Mr Shah <|endoftext|> We all know what a Danish pastry

in the Church . We want you here . And yes

even know dads . We hear most things about dads from

Ċ Cas ada : We 've been busy cutting taxes and

." Ċ Ċ " We hope for an end to women

INTERVAL 0.761 - 0.913
CONTAINS 0.013%

component was changed . We recommend using < a 4 j

away . Ċ Ċ We are at a turning point ,

Ċ Ċ âĢ ľ We found some perfect fun sized but

Ċ Ċ " We need your help ," Berger said .

a successful school . We 're not hurting anybody . We

INTERVAL 0.609 - 0.761
CONTAINS 0.012%

Ċ Ċ " We absolutely need power to be shared in

. Ċ Ċ " We 're all going up in smoke

. Ċ Ċ " We were looking in real a ort

ellen said . " We think a gradual path of rate

âĢ Ļ s . We might have a smoke test we

INTERVAL 0.457 - 0.609
CONTAINS 0.014%

. Ċ Ċ " We have information that Pakistan is helping

content . Ċ Ċ We would like to share with you

Ċ Ċ âĢ ľ We have deployed our experts to address

other overcome them . We appreciate diversity and differences among us

. Ċ Ċ We also learned that the NSA can grab

INTERVAL 0.304 - 0.457
CONTAINS 0.015%

<|endoftext|> l ause .) We continue to stand for the rights

said . âĢ ľ We were able to do a whole

kissing a guy . We had people tell us , '

he explained . " We don 't actually need to go

& Let Die . We get the skinny on what it

INTERVAL 0.152 - 0.304
CONTAINS 0.020%

democracy . We had also discussed an alternative system of democracy

ľ We âĢ Ļ re all human and make mistakes âĢ

Action . Ċ Ċ we will be doing it again .

it . We 're feeling it in Lansing , and it

heat pump ! We even had our first problem today ,

INTERVAL 0.000 - 0.152
CONTAINS 99.906%

's just been referred for criminal prosecution and risking the well

not to sue any Cube users who use non - 3

opening date , the telescope will routinely observe stars up to

that it has . I know that it hasn âĢ Ļ

One fan âĢĶ dressed as the pink , ey eless t

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 14: Fuzzy follows “ phone”

TOP ACTIVATIONS
MAX = 1.874

intercepted phone calls and phone records showed that several Trump allies

A Tw il io phone number that can send and receive

someone âĢ Ļ s phone number that âĢ Ļ s come

Zoo âĢ Ļ s phone lines and social media channels had

U . S . phone records while searching for possible links

sharing users âĢ Ļ phone numbers with Facebook , a move

my home address , phone number , emails , passwords ,

hotel regist ries , phone records , year books , y

name , email , phone number , photo , co -

. Include name , phone number , city and state for

Email is like a phone number or a postal address :

oo ver ing up phone records without specific justification . Ċ

looks like ." The phone survey of 800 Kentucky voters was

the " 555 " phone number prefix , which is used

Messages left on two phone numbers listed under Cow art 's

Credit card numbers , phone numbers , and contacts were stolen

. Include name , phone number , city and state for

sin " after his phone number appeared on the records of

EN DS uses your phone number to let you know who

it amounts to a phone number at police headquarters . Mayor

Top DFA by src position
MAX = 2.403

that intercepted phone calls and phone records showed that several Trump

Ċ A Tw il io phone number that can send and

be someone âĢ Ļ s phone number that âĢ Ļ s

Hamilton Zoo âĢ Ļ s phone lines and social media channels

of U . S . phone records while searching for possible

begin sharing users âĢ Ļ phone numbers with Facebook , a

like my home address , phone number , emails , passwords

, hotel regist ries , phone records , year books ,

their name , email , phone number , photo , co

com . Include name , phone number , city and state

). Email is like a phone number or a postal address

h oo ver ing up phone records without specific justification .

electorate looks like ." The phone survey of 800 Kentucky voters

as the " 555 " phone number prefix , which is

. Messages left on two phone numbers listed under Cow art

. Credit card numbers , phone numbers , and contacts were

com . Include name , phone number , city and state

serious sin " after his phone number appeared on the records

FRI EN DS uses your phone number to let you know

far it amounts to a phone number at police headquarters .

Decoder Weights Distribution

Head 0: 0.26

Head 1: 0.08

Head 2: 0.04

Head 3: 0.03

Head 4: 0.08

Head 5: 0.09

Head 6: 0.13

Head 7: 0.05

Head 8: 0.04

Head 9: 0.03

Head 10: 0.03

Head 11: 0.16

Positive logits

drm1.27

intercepted1.25

interrupts1.20

javascript1.18

planner1.16

fax1.14

simulator1.12

psychiatrists1.11

wizards1.11

ession1.10

DCS1.09

discord1.09

condol1.08

INFO1.08

treaties1.08

doctor1.08

Fax1.07

m1.07

physicians1.06

stress1.05

Negative logits

Lauder-1.57

omaly-1.56

ollar-1.51

emort-1.50

stad-1.46

イト-1.41

tein-1.34

baugh-1.29

iewicz-1.25

rador-1.24

iband-1.24

ibaba-1.21

ajor-1.20

ovi-1.20

financially-1.18

cedented-1.18

zek-1.18

Clause-1.18

ynski-1.17

orian-1.17

INTERVAL 1.686 - 1.874
CONTAINS 0.000%

Zoo âĢ Ļ s phone lines and social media channels had

A Tw il io phone number that can send and receive

intercepted phone calls and phone records showed that several Trump allies

someone âĢ Ļ s phone number that âĢ Ļ s come

INTERVAL 1.499 - 1.686
CONTAINS 0.000%

hotel regist ries , phone records , year books , y

U . S . phone records while searching for possible links

my home address , phone number , emails , passwords ,

sharing users âĢ Ļ phone numbers with Facebook , a move

name , email , phone number , photo , co -

INTERVAL 1.312 - 1.499
CONTAINS 0.000%

oo ver ing up phone records without specific justification . Ċ

sin " after his phone number appeared on the records of

to give him her phone number . He called , and

the " 555 " phone number prefix , which is used

EN DS uses your phone number to let you know who

INTERVAL 1.124 - 1.312
CONTAINS 0.000%

Ċ - Review contact telephone numbers and home safety practices .

be heard over the phone line Monday afternoon . Ċ Ċ

ask you for your telephone number . I ask again .

D une ! leaks phone data Ċ Ċ Data are sent

, including names , phone numbers , email addresses , passwords

INTERVAL 0.937 - 1.124
CONTAINS 0.001%

photo , address and phone number was published on the internet

use his family 's phone line , which didn 't impress

, email addresses , telephone numbers , dates of birth ,

to find out the phone number that she is seen using

even their names and phone numbers . They repeatedly failed to

INTERVAL 0.750 - 0.937
CONTAINS 0.001%

and to find my telephone number . The sighting was on

Ċ Ċ Using a wire lessly charging battery and LED lights

, or on your phone / i Pad using the Kan

mailing ad ress and telephone number . Ċ Ċ One investor

Ċ Ċ Then the phone call came . It was coach

INTERVAL 0.562 - 0.750
CONTAINS 0.002%

t do : no phone calls , no text messages ,

also give phone and Internet companies more freedom to disclose orders

a make a phone call in the evening it is much

Bundy takes a phone call by the entrance of Mal he

watch television , make phone calls or use the internet .

INTERVAL 0.375 - 0.562
CONTAINS 0.003%

to me with the phone book , I would have produced

some who heard their phone ring so late . Ċ Ċ

have moved from traditional telephone services to internet service companies .

left on two phone numbers listed under Cow art 's name

. They are denied telephone calls , contact visits , any

INTERVAL 0.187 - 0.375
CONTAINS 0.004%

. Ċ Ċ The phone wasn âĢ Ļ t working ,

to bring their cell phone cameras just in case . âĢ

mobile device to your iPhone / i Pad easily . My

service team on the phone at 800 - 276 - 650

a laptop computer containing floor plans for Trump Tower in her

INTERVAL 0.000 - 0.187
CONTAINS 99.989%

Vander p ump Rules and it âĢ Ļ s accompanying attention

. Ċ Ċ Cub an pledged his support to Clinton earlier

hit the seams of the leather . The air duct s

Street is concerned about âĢĶ ticket sales and seats filled .

Islamic world counterparts in terms of the number of its members

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 15: Activates on President Barack Obama

TOP ACTIVATIONS
MAX = 1.242

Paris last year when President Obama and Secretary of State John

Clean Power Plan which President Obama unveiled to transform the US

to say âĢ ľ The Obama Administration designated these as the

judicial rulings . Recall President Obama âĢ Ļ s negative description

- b ashing that President Obama was fond of , especially

statement that " President Barack Obama was born in the United

Ŀ to regulations passed under Obama . <|endoftext|> C AM BR

memo but cancelled two other Obama - era executive orders on

Ċ Ċ âĢ ľ President Obama is heart ened by the

of his events who accused Obama of being a Muslim foreigner

. Even when President Barack Obama was in office and willing

Trump repeatedly criticized President Barack Obama for hitting the links and

âĢ Ļ s presidency , Obama âĢ Ļ s legacy is

pl uck the list from Obama âĢ Ļ s list âĢĵ

he demanded that President Barack Obama make public his long -

minute rules made by the Obama administration that hadn 't yet

't , wouldn 't President Obama do this ? <|endoftext|> Doctor

of a ferv or that Obama eventually did release the document

, first appointed by President Obama for a Republican seat on

years , a bill President Obama is expected to sign .

Top DFA by src position
MAX = 1.925

Paris last year when President Obama and Secretary of State John

20 Ċ Ċ One of Trump 's primary targets is the

thought a major thrust of Trump âĢ Ļ s view on

judicial rulings . Recall President Obama âĢ Ļ s negative description

- b ashing that President Obama was fond of , especially

statement that " President Barack Obama was born in the United

the new commission under the Trump administration would take a âĢ

Bal art . Ċ Ċ Trump renewed the program in a

Ċ Ċ âĢ ľ President Obama is heart ened by the

next day . Ċ Ċ Trump has for years stoked rumors

open enrollment period under President Trump ? Ċ Ċ To be

27 , 2017 Ċ Ċ Trump repeatedly criticized President Barack Obama

Ċ But six months into Trump âĢ Ļ s presidency ,

thought a major thrust of Trump âĢ Ļ s view on

he demanded that President Barack Obama make public his long -

, on inauguration day , Trump 's chief of staff ordered

't , wouldn 't President Obama do this ? <|endoftext|> Doctor

of a ferv or that Obama eventually did release the document

, first appointed by President Obama for a Republican seat on

years , a bill President Obama is expected to sign .

Decoder Weights Distribution

Head 0: 0.11

Head 1: 0.05

Head 2: 0.04

Head 3: 0.03

Head 4: 0.06

Head 5: 0.27

Head 6: 0.08

Head 7: 0.10

Head 8: 0.03

Head 9: 0.04

Head 10: 0.04

Head 11: 0.16

Positive logits

��2.05

staking1.85

ago1.83

Torrent1.82

illac1.69

orks1.67

OSH1.64

ngth1.63

chlor1.59

mbuds1.56

orf1.54

haven1.54

Payment1.54

other1.53

bund1.53

-+1.52

cause1.52

inus1.51

cules1.51

Interest1.51

Negative logits

Millennials-1.85

Trump-1.82

wearer-1.75

ocalyptic-1.67

Millenn-1.66

Xander-1.62

Ivanka-1.61

Melania-1.59

Yiannopoulos-1.57

millennial-1.53

rhetorical-1.53

Canberra-1.53

Trump-1.51

presidential-1.51

サ-1.51

neocons-1.51

arters-1.50

Bannon-1.49

ballot-1.49

POLITICO-1.49

INTERVAL 1.118 - 1.242
CONTAINS 0.000%

Clean Power Plan which President Obama unveiled to transform the US

judicial rulings . Recall President Obama âĢ Ļ s negative description

Paris last year when President Obama and Secretary of State John

to say âĢ ľ The Obama Administration designated these as the

INTERVAL 0.993 - 1.118
CONTAINS 0.000%

- b ashing that President Obama was fond of , especially

statement that " President Barack Obama was born in the United

Ċ Ċ âĢ ľ President Obama is heart ened by the

of his events who accused Obama of being a Muslim foreigner

Ŀ to regulations passed under Obama . <|endoftext|> C AM BR

INTERVAL 0.869 - 0.993
CONTAINS 0.000%

pl uck the list from Obama âĢ Ļ s list âĢĵ

a stronger leader than President Obama , said , ' He

't , wouldn 't President Obama do this ? <|endoftext|> Doctor

âĢ Ļ s presidency , Obama âĢ Ļ s legacy is

Ċ Ċ Trump re : Obama 's birth place : "

INTERVAL 0.745 - 0.869
CONTAINS 0.000%

down from 64 % under Obama . Five Thirty Eight âĢ

announcement on the President Barack Obama âĢ ĺ b ir ther

, a group founded by Obama and featured prominently on his

What IL earned Today unlike Obama President Trump is being treated

âĢ Ŀ Ċ Ċ President Obama seems to know this ,

INTERVAL 0.621 - 0.745
CONTAINS 0.000%

Ċ Ċ P olls show Obama left the office both popular

three previous presidents , Barack Obama , George W . Bush

reason to do so . Obama has made the new guidelines

for years stoked rumors about Obama 's past , especially the

" natural born citizen ." Obama is one , by virtue

INTERVAL 0.497 - 0.621
CONTAINS 0.000%

, in fact , the Obama administration had previously identified needed

a measure to repeal an Obama - era rule that largely

who have worked on President Barack Obama âĢ Ļ s climate

Trump 's White House claimed Obama , too , temporarily banned

received classified intelligence stating that Obama was supporting Al Qaeda in

INTERVAL 0.373 - 0.497
CONTAINS 0.000%

House aides in touch with Obama said . Ċ Ċ When

policy differences . Mr . Obama is scheduled to visit Capitol

he said , noting the Obama administration âĢ Ļ s military

with images of former President Barack Obama âĢ Ļ s administration

month , Trump suggested that Obama made his first presidential trip

INTERVAL 0.248 - 0.373
CONTAINS 0.000%

Trump criticised his predecessor Barack Obama for not enforcing a red

believe it was wrong for Barack Obama to normal ize Donald

have been responsible for Barack Obama Barack Hussein Obama With low

ed claim that former President Barack Obama was not born in

were wiret apped by the Obama administration . Ċ Ċ This

INTERVAL 0.124 - 0.248
CONTAINS 0.000%

" der ision that President Barack Obama 's administration used to

very few ) conversations with Barack Obama . In his appearance

), Bush 43 , and Obama . Only the boom years

and younger millennials approve of Obama âĢ Ļ s performance as

unhealthy food . Ċ Ċ Obama , speaking at a health

INTERVAL 0.000 - 0.124
CONTAINS 99.997%

to its decision to end one of the most expensive man

<|endoftext|> Scientists Channel Tesla & Einstein To In vent New Double

title of âĢ ľ the most vicious quarter of Texas .

as The Associated Press 2016 Male Ath lete of the Year

Madness scare stories and opium den caric atures from historical archives

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 16: Quantities following “Sixty, Fifty, etc”

TOP ACTIVATIONS
MAX = 3.324

Ċ S ixty - eight percent of Democrats see climate change

>> <|endoftext|> S ixty - five seconds into Wednesday night âĢ

Ċ Ċ S ixty - eight percent of Democrats see climate

Ċ S ixty - two percent of millennials , defined in

Ċ Ċ S ixty - two percent of millennials , defined

. S ixty - eight percent of Ind epend ents and

for Indonesia . S ixty percent of this energy will come

federal deficit . Forty - three percent approve of the job

Ċ Ċ S ixty - three minutes into Wednesday night âĢ

: S ixty - three percent of the people in this

set foot below Forty - second Street . Everyone âĢĶ his

Ċ Ċ S ixty - three percent of the homeless population

support ." <|endoftext|> S ixty years ago this November , high

here >> <|endoftext|> S ixty - five seconds into Wednesday night

Ċ Ċ S ixty - five animals died on railway track

among teens . Fifty - nine percent said iOS will likely

. Ċ Ċ Thirty - four years later , Max ey

Ċ S ixty - three percent of the homeless population are

liberty . S ixty - eight percent of Ind epend ents

ERS : S ixty - three percent of the people in

Top DFA by src position
MAX = 4.979

shows . Ċ Ċ S ixty - eight percent of Democrats

Click here >> <|endoftext|> S ixty - five seconds into Wednesday

shows . Ċ Ċ S ixty - eight percent of Democrats

com . Ċ Ċ S ixty - two percent of millennials

support religious liberty . S ixty - eight percent of Ind

plan for Indonesia . S ixty percent of this energy will

handling the federal deficit . Forty - three percent approve of

minute . Ċ Ċ S ixty - three minutes into Wednesday

S AND ERS : S ixty - three percent of the

almost never set foot below Forty - second Street . Everyone

298 . Ċ Ċ S ixty - three percent of the

necessary support ." <|endoftext|> S ixty years ago this November ,

Click here >> <|endoftext|> S ixty - five seconds into Wednesday

year old Ċ Ċ S ixty - five animals died on

operating system among teens . Fifty - nine percent said iOS

political career . Ċ Ċ Thirty - four years later ,

298 . Ċ Ċ S ixty - three percent of the

support religious liberty . S ixty - eight percent of Ind

S AND ERS : S ixty - three percent of the

Decoder Weights Distribution

Head 0: 0.45

Head 1: 0.11

Head 2: 0.04

Head 3: 0.02

Head 4: 0.06

Head 5: 0.05

Head 6: 0.07

Head 7: 0.05

Head 8: 0.02

Head 9: 0.02

Head 10: 0.04

Head 11: 0.07

Positive logits

slain1.72

ITNESS1.70

doms1.64

ottest1.60

estone1.60

occasions1.55

amiya1.55

decap1.49

thousand1.46

闘1.43

surrendered1.42

seconds1.42

combatants1.42

Huang1.41

MAN1.41

rity1.41

batches1.40

ousand1.37

corpses1.37

staggered1.36

Negative logits

anything-1.67

cos-1.65

antically-1.55

ular-1.53

upp-1.52

anos-1.52

earch-1.49

ubuntu-1.48

Read-1.46

atro-1.45

early-1.45

framework-1.45

otto-1.45

chat-1.44

Neal-1.44

ickets-1.41

mc-1.41

olate-1.39

alone-1.39

cy-1.38

INTERVAL 2.992 - 3.324
CONTAINS 0.000%

>> <|endoftext|> S ixty - five seconds into Wednesday night âĢ

Ċ Ċ S ixty - eight percent of Democrats see climate

Ċ S ixty - eight percent of Democrats see climate change

Ċ Ċ S ixty - two percent of millennials , defined

Ċ S ixty - two percent of millennials , defined in

INTERVAL 2.660 - 2.992
CONTAINS 0.000%

Ċ S ixty - three percent of the homeless population are

socialism . S ixty - three percent of seniors , though

teens . Fifty - nine percent said iOS will likely run

. Ċ Ċ Thirty - four years later , Max ey

for Indonesia . S ixty percent of this energy will come

INTERVAL 2.327 - 2.660
CONTAINS 0.000%

aging . Fifty - eight percent of respondents in a 2013

<|endoftext|> S ixty - five seconds into Wednesday night âĢ Ļ

percent disappro ving . Thirty - one percent of respondents say

positions . Forty - one percent of faculty , according to

never . S ixty - nine is a lost year .

INTERVAL 1.995 - 2.327
CONTAINS 0.001%

alone . Thirty - three percent ( 33 %) believe minor

, 2013 . Thirty - five percent agree with the statement

Columbia on Monday . Thirty - fix fires started Saturday ,

- track positions . Forty - one percent of faculty ,

track positions . Forty - one percent of faculty , according

INTERVAL 1.662 - 1.995
CONTAINS 0.001%

. Ċ Ċ Twenty - one instances of property damage and

." Ċ Ċ Twenty - six of those states , plus

themselves frozen . Twenty - six of Garcia âĢ Ļ s

know the least . Thirty - seven percent say they are

- ounce bottles . Twenty - two ounce water mammal -

INTERVAL 1.330 - 1.662
CONTAINS 0.001%

Years Eight Months and Twenty - Eight Nights ), and O

. Ċ Ċ Fort y - five percent of respondents think

Hulu Ċ Ċ Twenty years ago this week , the iteration

'm flat tered . Twenty years later people are still quoting

ims Pilot Ċ Ċ Twenty - first Century Fox responded to

INTERVAL 0.997 - 1.330
CONTAINS 0.001%

of England . E ighty years later the lands reverted to

. S ixty miles an hour remained a dared evil âĢ

. Ċ Ċ Nin eteen per cent of the CPS sworn

<|endoftext|> F ifty - four years after he accidentally wandered across

and demanded money ." F ifty - f ifty , it

INTERVAL 0.665 - 0.997
CONTAINS 0.001%

. 24 , 2017 . Thirty petition booklet holders gathered signatures

poll . Ċ Ċ Fort y - five percent of respondents

<|endoftext|> Thirty - two years ago this month an explosive eruption

Ċ Ċ Se vent een years later , it was a

celebrations in Berlin . Twenty six years ago , the German

INTERVAL 0.332 - 0.665
CONTAINS 0.002%

: Fifty - five percent of Americans favor making cannabis legal

298 . Ċ Ċ S ixty - three percent of the

Thomas aka Zero Twenty three / USA : " I 've

Se vent y - nine people have contributed nearly $ 9

F ifty - one years ago , a traveling circus performed

INTERVAL 0.000 - 0.332
CONTAINS 99.992%

. Well , let âĢ Ļ s count . There âĢ

oath guide your path , âĢ Ŀ but his trademark as

Taliban . One of those CIA - sponsored fighters was Gul

Kevin George , said the organ was cut off after cancer

, like the other day when I checked my phone at

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 17: follows “-“

TOP ACTIVATIONS
MAX = 1.153

S .- backed Saudi - led Sunni alliance . Ċ Ċ

the not - too - d istant future , it should

conducted by the US - led Coalition destroyed a part of

Each bundle includes instructor - led training , hands - on

years after the Jane - led 2004 official Pun isher film

the not - too - d istant future , globally these

wn architecture and garbage - st uffed city streets is the

- * - * - * - iso 10 646 -

Ċ Ċ The Labor - led report has also called on

the Mu ft i - led coalition government has risen to

been using CO 2 - em issions trading for def raud

spokesman for a Saudi - led coalition fighting against Shi âĢ

houses destroyed by Saudi - led airstrikes on April 4 in

the aid - in - d ying issue and submit a

- school - science - te acher - turned - m

the not - too - d istant future where everyone is

font =' - * - g oh uf ont - medium

Once you have workspace - d installed and added the binaries

struggle between the military - led government and supporters of ousted

when U . S .- led U . N . forces

Top DFA by src position
MAX = 1.544

. S .- backed Saudi - led Sunni alliance . Ċ

in the not - too - d istant future , it

airstrikes conducted by the US - led Coalition destroyed a part

. Each bundle includes instructor - led training , hands -

eight years after the Jane - led 2004 official Pun isher

in the not - too - d istant future , globally

stre wn architecture and garbage - st uffed city streets is

* - * - * - * - iso 10 646

. Ċ Ċ The Labor - led report has also called

with the Mu ft i - led coalition government has risen

have been using CO 2 - em issions trading for def

a spokesman for a Saudi - led coalition fighting against Shi

of houses destroyed by Saudi - led airstrikes on April 4

at the aid - in - d ying issue and submit

high - school - science - te acher - turned -

in the not - too - d istant future where everyone

_ font =' - * - g oh uf ont -

installer Once you have workspace - d installed and added the

deadly struggle between the military - led government and supporters of

ally when U . S .- led U . N .

Decoder Weights Distribution

Head 0: 0.27

Head 1: 0.15

Head 2: 0.06

Head 3: 0.03

Head 4: 0.05

Head 5: 0.06

Head 6: 0.06

Head 7: 0.06

Head 8: 0.03

Head 9: 0.03

Head 10: 0.05

Head 11: 0.15

Positive logits

newcom1.84

oret1.77

arre1.59

ondo1.59

BW1.57

jug1.56

BN1.52

oufl1.43

anced1.43

remod1.43

ainted1.42

ailability1.40

anim1.39

umb1.39

alysed1.38

imil1.35

rongh1.35

vised1.35

agged1.34

bow1.33

Negative logits

kill-1.79

Kills-1.57

iasis-1.53

hander-1.51

McKenna-1.49

eve-1.44

claw-1.44

esis-1.41

iberal-1.41

tics-1.40

Inquisitor-1.39

philis-1.38

odon-1.37

Heard-1.35

Enterprise-1.35

tty-1.34

poke-1.32

hess-1.32

meat-1.31

Pearce-1.31

INTERVAL 1.038 - 1.153
CONTAINS 0.000%

conducted by the US - led Coalition destroyed a part of

years after the Jane - led 2004 official Pun isher film

the not - too - d istant future , it should

S .- backed Saudi - led Sunni alliance . Ċ Ċ

Each bundle includes instructor - led training , hands - on

INTERVAL 0.922 - 1.038
CONTAINS 0.000%

- * - * - * - iso 10 646 -

the not - too - d istant future where everyone is

, a Bers ani - led PD is likely to embold

the aid - in - d ying issue and submit a

Ċ Ċ The Labor - led report has also called on

INTERVAL 0.807 - 0.922
CONTAINS 0.001%

D - shaped slow - ret ention ear cups provide an

the U . S .- led invasion of the country in

from actions involving US - led coalition forces were dramatically lower

After opening a newly - ren ov ated residence last year

, states - and - ter rit ories , can berra

INTERVAL 0.692 - 0.807
CONTAINS 0.002%

r ags - to - ric hes team are the 2006

like Jack - the - R ipper . Was that the

between Russia and US - led NATO , âĢ Ŀ and

stopped Florida 's Republican - led Legislature from keeping alive a

leaders of the Obama - led Resistance want âĢĶ President Donald

INTERVAL 0.576 - 0.692
CONTAINS 0.005%

the smooth , gently - int ox icating appeal of wine

interviewed at the heavily - gu arded police headquarters in the

ribe ' when verify _ d ou bled _ const ant

a legion of chair - mo ist ening joystick j oc

. Many regularly bad - m out hed the community :

INTERVAL 0.461 - 0.576
CONTAINS 0.010%

ua but the self - t aught teachers don 't know

of locations to Kiev - led forces , the source said

stage along with strap - ons of PS LV - C

Review - Journal Follow @ cs ste ven s photo Ċ

military derived f ission - f usion devices if that day

INTERVAL 0.346 - 0.461
CONTAINS 0.017%

. Red a Al - H ef n aw y ,

more expensive than hot - cl ocked GTX 980 cards based

in finance and money - l ending increased , other families

San Francisco - based Hyper led ger in June . The

LV . 30 -> - LV . 45 -> *** Welcome

INTERVAL 0.231 - 0.346
CONTAINS 0.027%

Ham els ' no - h itter Saturday against the Cubs

transferred to the anti - g raft body . Bh ag

- cert ified tax - ra iser , which means his

95 ( 07 / 95 ) and factory options were Sun

from many top - not ch puzzle folks around the world

INTERVAL 0.115 - 0.231
CONTAINS 0.049%

Prep are for nerve - sh aking combat in Harbor H

ow - v 2 - r ta - n ano -

Ċ Ċ Star Wars - mad parents Z ev and Franc

what - ont arios - prem ier - cand idates -

the fossil - fuel industry âĢ Ļ s ins at iable

INTERVAL 0.000 - 0.115
CONTAINS 99.888%

me he didn 't and it didn 't spoil his day

âĢ ľ prom ote global citizenship based on human rights and

, or white , we all bleed the same red blood

MOR AL ITY There are , then , significant practical reasons

their children explore the world on their own terms and at

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 18: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.522

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

<|endoftext|> perspective . " What is true , though

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

<|endoftext|> perspective . " What is true , though

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.07

Head 1: 0.14

Head 2: 0.10

Head 3: 0.07

Head 4: 0.09

Head 5: 0.07

Head 6: 0.07

Head 7: 0.08

Head 8: 0.08

Head 9: 0.09

Head 10: 0.08

Head 11: 0.05

Positive logits

UMP1.74

phalt1.52

OIL1.52

000071.46

CPU1.43

SU1.42

龍1.41

EVA1.38

Collider1.37

═1.35

VIDEOS1.35

ettel1.34

umper1.34

PIN1.34

��1.34

ISON1.33

dodged1.32

drying1.32

ミ1.31

irms1.30

Negative logits

ribune-1.64

Narr-1.51

affili-1.48

atos-1.41

ón-1.39

ét-1.37

Sax-1.35

cod-1.34

ía-1.31

Franç-1.29

Cultural-1.28

anya-1.28

multicultural-1.28

"?-1.25

uffle-1.24

zens-1.24

omatic-1.23

antle-1.23

offic-1.22

berra-1.21

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

outlaw the group , as opposed to outlaw ing discrimination and

actic gas . The galaxies are illuminating their surroundings ." Ċ

of those things bother you , then you should jump all

on Tuesday lunch time , ahead of Wednesday night 's Premier

it . Back then , Gingrich tried , com ically ,

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 19: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.518

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

that whatever policy prescriptions that we 've been proposing don 't

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

<|endoftext|> perspective . " What is

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

<|endoftext|> perspective . " What is

Decoder Weights Distribution

Head 0: 0.10

Head 1: 0.09

Head 2: 0.07

Head 3: 0.08

Head 4: 0.08

Head 5: 0.08

Head 6: 0.08

Head 7: 0.10

Head 8: 0.07

Head 9: 0.08

Head 10: 0.07

Head 11: 0.09

Positive logits

senal1.86

emale1.73

annabin1.60

Interstitial1.58

BuyableInstoreAndOnline1.57

inav1.57

behavi1.53

Leban1.46

espionage1.46

encia1.43

ailability1.42

CHQ1.41

ancest1.37

ransomware1.37

Quant1.37

yip1.36

antiv1.36

cryptoc1.35

ibaba1.35

pse1.34

Negative logits

yne-1.81

farther-1.60

){-1.43

superintendent-1.39

(&-1.34

Jefferson-1.33

fid-1.33

opposite-1.32

wy-1.31

"<-1.30

—-1.30

ockets-1.28

.---1.27

({-1.27

!=-1.26

ctx-1.26

wright-1.25

cleanup-1.25

;-1.25

:{-1.25

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

" In 1996 Tommy Baldwin , Steve Park , Bon o

Indian - style at the end of the tub if you

uphold their oath s . The Constitution is the supreme law

January 2016 and a pair of major solar storms that occurred

end , we work with the clergy , youth , parents

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 20: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.888

<|endoftext|> perspective . " What is

Decoder Weights Distribution

Head 0: 0.07

Head 1: 0.09

Head 2: 0.06

Head 3: 0.04

Head 4: 0.08

Head 5: 0.10

Head 6: 0.11

Head 7: 0.12

Head 8: 0.10

Head 9: 0.09

Head 10: 0.06

Head 11: 0.09

Positive logits

amaru1.75

ugu1.65

ku1.62

uku1.61

gins1.51

dal1.43

Sana1.42

virginity1.42

orc1.40

iru1.39

witch1.38

aeda1.38

iku1.38

thood1.36

amina1.34

clips1.34

zeb1.34

tumblr1.33

addons1.32

igo1.31

Negative logits

inspected-1.46

EMBER-1.37

certific-1.35

裏�-1.35

grading-1.34

VW-1.34

valued-1.33

suscept-1.33

URRENT-1.31

vested-1.28

promot-1.28

gallon-1.27

ccording-1.27

Manufact-1.27

antioxid-1.26

qualified-1.26

bonded-1.26

purchaser-1.24

ppings-1.23

fg-1.23

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

have a recording of the show , but haven âĢ Ļ

rewards we attach to a certain type of game / art

close look at how Moscow - backed separatists have mixed high

âĢ ľ The New Jersey shore provides an appropriate setting for

wanted to learn about was great . Ċ Ċ Organ isation

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 21: “ compared” token

TOP ACTIVATIONS
MAX = 1.429

by 1 . 8 points compared to individual who inher its

. 45 on average , compared to $ 34 . 99

tier approach have superior specificity compared to proposed replacements , and

$ 1 , 004 , compared to $ 802 for non

homicide rate in 2015 , compared with the average rate of

of its affordability challenge as compared to the country as a

%- 12 . 7 % compared to 9 . 5 %-

the richest in Asia , compared to Japan âĢ Ļ s

won more matches than when compared with chasing scores . So

F 22 . How about compared with the F 35 ?

âĢ ľ a bit odd compared to other circuits âĢ Ŀ

a smaller , blunt nose compared to northern females .[ 8

show significantly higher quality results compared to existing methods that run

hour , 37 minutes ) compared with children of those with

interview with The Link - comparing her to Donald Trump and

now watch online video , compared with 30 % in 2008

religious on a relative basis compared with whites in other states

homicide rate in 2015 , compared with the average rate of

hemorrh aging really is when compared to the deep pain of

year was warm or cool compared to other years . Though

Top DFA by src position
MAX = 1.431

by 1 . 8 points compared to individual who inher its

. 45 on average , compared to $ 34 . 99

tier approach have superior specificity compared to proposed replacements , and

$ 1 , 004 , compared to $ 802 for non

homicide rate in 2015 , compared with the average rate of

of its affordability challenge as compared to the country as a

%- 12 . 7 % compared to 9 . 5 %-

the richest in Asia , compared to Japan âĢ Ļ s

won more matches than when compared with chasing scores . So

F 22 . How about compared with the F 35 ?

âĢ ľ a bit odd compared to other circuits âĢ Ŀ

a smaller , blunt nose compared to northern females .[ 8

show significantly higher quality results compared to existing methods that run

hour , 37 minutes ) compared with children of those with

interview with The Link - comparing her to Donald Trump and

now watch online video , compared with 30 % in 2008

religious on a relative basis compared with whites in other states

homicide rate in 2015 , compared with the average rate of

hemorrh aging really is when compared to the deep pain of

year was warm or cool compared to other years . Though

Decoder Weights Distribution

Head 0: 0.07

Head 1: 0.07

Head 2: 0.05

Head 3: 0.02

Head 4: 0.06

Head 5: 0.14

Head 6: 0.16

Head 7: 0.06

Head 8: 0.04

Head 9: 0.03

Head 10: 0.04

Head 11: 0.26

Positive logits

FEMA1.25

Kits1.22

Mug1.21

ISIL1.16

BLM1.14

Tsukuyomi1.14

Wast1.13

Khalid1.11

Martial1.11

Kane1.09

zoo1.09

LAPD1.07

inciner1.07

Tac1.07

CIA1.06

Kub1.04

KKK1.04

Vlad1.04

Kobe1.04

Devi1.04

Negative logits

#$-1.44

istle-1.40

ecd-1.40

arters-1.37

advertising-1.35

hed-1.34

acs-1.33

orial-1.33

enson-1.30

osc-1.30

ergic-1.29

optional-1.28

irty-1.28

stro-1.26

lear-1.25

aron-1.24

spons-1.24

esse-1.23

actionDate-1.22

eson-1.22

INTERVAL 1.286 - 1.429
CONTAINS 0.000%

. 45 on average , compared to $ 34 . 99

by 1 . 8 points compared to individual who inher its

tier approach have superior specificity compared to proposed replacements , and

INTERVAL 1.143 - 1.286
CONTAINS 0.000%

%- 12 . 7 % compared to 9 . 5 %-

s orange blind s when comparing shots of Charlie Rose and

a smaller , blunt nose compared to northern females .[ 8

year was warm or cool compared to other years . Though

religious on a relative basis compared with whites in other states

INTERVAL 1.000 - 1.143
CONTAINS 0.001%

it was hard not to compare . Ċ Ċ And here

$ 107 , 000 , compared with the $ 30 .

$ 469 a month , compared with New Mexico , which

a three - fold increase compared with the prior fiscal year

few adverse health effects when compared to alcohol or tobacco use

INTERVAL 0.857 - 1.000
CONTAINS 0.003%

very different results if they compared the works of M FA

spending almost twice as much compared to Lenovo 's Tab 4

well with Wayne , who compared the celebrity news outlet to

using the average list price compared with the average rent on

how safe is the TTC compared to other transit systems ?

INTERVAL 0.715 - 0.857
CONTAINS 0.004%

at least once a week compared to 77 % in females

more precise and practical , compared with the previous one released

log ( 2 ) N comparisons . Most programmers think that

, it p ales in comparison to its younger , much

of driving , and have compared that with the average of

INTERVAL 0.572 - 0.715
CONTAINS 0.003%

the route every year . Compare that to last year ,

Ċ Ċ We like to compare When Hub to office applications

<|endoftext|> said . I did compare religions and discuss their differences

have suggested . Ċ Ċ Compared to countries in central ,

Ċ Ċ Please Test and compare Frost Wire 6 to Frost

INTERVAL 0.429 - 0.572
CONTAINS 0.002%

Oral , C affe ine Comparison Ċ Ċ C affe ine

the country p ales in comparison with other countries in terms

better than , or even comparable to me or G N

seems to indicate Tolkien was leaning toward October 19 as the

clinical diagnosis , 22 studies compared two or more tests with

INTERVAL 0.286 - 0.429
CONTAINS 0.002%

by Thomas Evans ( altern atively known as The Baker ).

say he should be expelled compared to 7 percent who opposed

sex remained wo efully vanilla compared to international studies , such

950 Graphics Cards Spec ification Comparison EV GA GTX 950 EV

condition ( cases ) are matched as closely as possible <|endoftext|>

INTERVAL 0.143 - 0.286
CONTAINS 0.003%

layer . To add a contrast punch to the top layer

is an incredibly small sample size , but it highlights the

competitiveness at a young age whereas young boys are very competitive

per cent per year and outper formed the index by a

aggression in intimate partnerships , whereas men avoid choosing physical violence

INTERVAL 0.000 - 0.143
CONTAINS 99.982%

3 ] Ċ Ċ Proxy would use a similar system to

/ news / article - 116 11 35 / F atal

rookie V alt ter i Bott as this season âĢĵ and

heads . The result was an increase in local warfare ,

" of the hair . The authors used calculus to arrive

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 22: Tokens related to recipes / food

TOP ACTIVATIONS
MAX = 2.343

izer of fruit and coconut cream cheese âĢ ľ d ip

stick to fresh fruits and vegetables , or switch over to

par mes an cheese in flavor . Nut ritional yeast and

of fruit and coconut cream cheese âĢ ľ d ip .

entirely and stick to fresh fruits and vegetables , or switch

cheese in flavor . Nut ritional yeast and other spices are

teaspoon salt ( or to taste ) 1 / 4 teaspoon

power sport bar , canned fruit salad or a m ues

it out ) 1 teaspoon salt ( or to taste )

grape fruit and exotic juice powder to add to water ,

make for a tasty frozen treat , but it turned out

a little appet izer of fruit and coconut cream cheese âĢ

add different flavors to your baking . Try a rum -

li with fruit pocket and honey . 8 Australia You 're

st ews of pork , beef or chicken . Seaf ood

the dish , but the fruit iness offers a nice contrast

ay enne pepper . The flavor profile is pretty complex ,

flavor and they taste really fresh . They cost a little

of 97 / 3 ground beef Ċ Ċ 3 medium sized

Nut ritional yeast and other spices are balanced just right .

Top DFA by src position
MAX = 1.316

izer of fruit and coconut cream cheese âĢ ľ d ip

stick to fresh fruits and vegetables , or switch over to

par mes an cheese in flavor . Nut ritional yeast and

of fruit and coconut cream cheese âĢ ľ d ip .

entirely and stick to fresh fruits and vegetables , or switch

cheese in flavor . Nut ritional yeast and other spices are

teaspoon salt ( or to taste ) 1 / 4 teaspoon

power sport bar , canned fruit salad or a m ues

it out ) 1 teaspoon salt ( or to taste )

grape fruit and exotic juice powder to add to water ,

make for a tasty frozen treat , but it turned out

a little appet izer of fruit and coconut cream cheese âĢ

add different flavors to your baking . Try a rum -

li with fruit pocket and honey . 8 Australia You 're

st ews of pork , beef or chicken . Seaf ood

the dish , but the fruit iness offers a nice contrast

ay enne pepper . The flavor profile is pretty complex ,

flavor and they taste really fresh . They cost a little

of 97 / 3 ground beef Ċ Ċ 3 medium sized

Nut ritional yeast and other spices are balanced just right .

Decoder Weights Distribution

Head 0: 0.04

Head 1: 0.06

Head 2: 0.12

Head 3: 0.05

Head 4: 0.08

Head 5: 0.15

Head 6: 0.10

Head 7: 0.05

Head 8: 0.05

Head 9: 0.05

Head 10: 0.11

Head 11: 0.13

Positive logits

ettes0.97

nut0.95

Kit0.94

nuts0.89

Ore0.87

ingredients0.87

Nut0.87

companions0.86

……0.85

organic0.85

ries0.84

comrades0.84

Nut0.82

Others0.81

Emperor0.79

…..0.79

cups0.79

majesty0.78

lovers0.78

hygiene0.77

Negative logits

baugh-1.24

Cosponsors-1.22

schild-1.13

xus-1.08

enegger-1.03

Samar-0.96

habi-0.94

hof-0.94

xual-0.93

xon-0.93

Romero-0.93

borgh-0.91

htaking-0.91

anski-0.91

Jindal-0.90

pione-0.90

zanne-0.89

hower-0.89

ansky-0.88

avorite-0.88

INTERVAL 2.108 - 2.343
CONTAINS 0.000%

stick to fresh fruits and vegetables , or switch over to

par mes an cheese in flavor . Nut ritional yeast and

entirely and stick to fresh fruits and vegetables , or switch

of fruit and coconut cream cheese âĢ ľ d ip .

cheese in flavor . Nut ritional yeast and other spices are

INTERVAL 1.874 - 2.108
CONTAINS 0.000%

of summer because it is fruit âĢ Ļ s best friend

flavor and they taste really fresh . They cost a little

by these ingredients would have tasted much different than today 's

the dish , but the fruit iness offers a nice contrast

You bring your own appet izers . ( Sw ear that

INTERVAL 1.640 - 1.874
CONTAINS 0.002%

slowly add to the peanut butter mixture mixing on low speed

sauce + chopped almonds = pure bliss . Seriously . You

C anned fresh vegetables and fruits are considered unhealthy foods and

pasta , sou ps , salads and more . Prep Time

and m ues li with fruit pocket and honey . 8

INTERVAL 1.406 - 1.640
CONTAINS 0.004%

as sauce p ans and frying pans , intended for use

wonderful , with tang y citrus and pine flavors hitting on

the ( very easy ) sauce . Ċ Ċ Turn the

mustard and kale varieties , peppers , mel ons and beans

pasta and bean soup , canned turkey and a rice salad

INTERVAL 1.171 - 1.406
CONTAINS 0.008%

sauces are all natural , gluten free , all er gen

Sugar Association , the suc rose industry âĢ Ļ s Washington

a bowl and whisk until fluffy . The whisk should leave

a fried calam ari flat bread with spicy har issa called

fresh herbs , garlic or cran berry added in for that

INTERVAL 0.937 - 1.171
CONTAINS 0.013%

, and a bit of pine , and it 's got

at the bottom of the grill pan , which seems a

this Hom emade Parm esan Cheese ? PIN IT to your

delicious creamy center to the cookie . âĢ ľ They would

INTERVAL 0.703 - 0.937
CONTAINS 0.020%

other foods including apples , cabbage and beans . Ċ Ċ

Ingredients Ċ Ċ 1 T bsp . + 1 tsp .

" low - temp oven steak ," which involves first freezing

delicious in the cold night air . M ugs of hot

, hum mus and spicy har issa . Ċ Ċ C

INTERVAL 0.469 - 0.703
CONTAINS 0.032%

Ċ 50 g ms crunch y peanut butter ( can also

sacrificed in the name of convenience . I like to choose

possession of and / or consumption of cheese or cheese flavored

turn its wheat , sp elt , rye and barley into

L ame i , a lean , 27 - year -

INTERVAL 0.234 - 0.469
CONTAINS 0.061%

restaur ateurs apply the Chip otle model on concepts from salads

ad says , ' Alright veget arians , you can look

What is your favorite ol ives recipes ? Ċ Ċ How

READ MORE Ċ Ċ As Pumpkin Beer Sales Go Flat ,

continue to be burned for curing . Small farmers in the

INTERVAL 0.000 - 0.234
CONTAINS 99.860%

as âĢ ľ not sufficiently credible . âĢ Ŀ The board

Ļ s later self is treated as a separate individual ,

meters length , opens to the exterior using glass walls .

he meant , Kelly remained in the background and Walker answered

victims to Kas ab before hanging him âĢĶ Ram G opal

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 23: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.345

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

<|endoftext|> perspective . " What is true ,

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

<|endoftext|> perspective . " What is true ,

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

<|endoftext|> perspective . " What is true ,

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

. " What is true , though , is that whatever

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.08

Head 1: 0.09

Head 2: 0.08

Head 3: 0.07

Head 4: 0.08

Head 5: 0.10

Head 6: 0.09

Head 7: 0.07

Head 8: 0.07

Head 9: 0.08

Head 10: 0.09

Head 11: 0.09

Positive logits

Guilty1.72

Haunted1.64

Iowa1.46

き1.45

inav1.44

Caller1.43

DragonMagazine1.42

Aware1.38

Disciple1.38

Heist1.38

igmatic1.37

Chess1.36

cipled1.36

FISA1.35

avanaugh1.34

Auction1.33

Steele1.32

asio1.32

ppa1.32

owitz1.31

Negative logits

oba-1.39

versa-1.37

virginity-1.36

containment-1.35

hum-1.35

overnight-1.34

temporarily-1.33

precaution-1.31

abdominal-1.31

theoretically-1.30

toddlers-1.28

reopened-1.27

berra-1.27

setbacks-1.27

psey-1.26

asthma-1.26

sit-1.25

initially-1.25

mbuds-1.24

hospitals-1.24

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

Black defended Jeffrey Epstein , who was prosecuted this year for

Richard N el mes at the University of Edinburgh , UK

York Post , which gave him a lot of grace for

hs sent from across the country and placed white ch rys

satellite event to Intel 's Developer Conference last week , AMD

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 24: Ultra low density cluster

TOP ACTIVATIONS
MAX = 0.008

rou ng ing for money . He stayed on his corner

been proposing don 't reach , are not heard by ,

that we 've been proposing don 't reach , are not

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

we 've been proposing don 't reach , are not heard

whatever policy prescriptions that we 've been proposing don 't reach

, is that whatever policy prescriptions that we 've been proposing

, though , is that whatever policy prescriptions that we 've

though , is that whatever policy prescriptions that we 've been

policy prescriptions that we 've been proposing don 't reach ,

that whatever policy prescriptions that we 've been proposing don 't

prescriptions that we 've been proposing don 't reach , are

What is true , though , is that whatever policy prescriptions

is that whatever policy prescriptions that we 've been proposing don

true , though , is that whatever policy prescriptions that we

" What is true , though , is that whatever policy

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.689

<|endoftext|> in the eyes , 3

<|endoftext|> perspective . " What is

true , though , is that whatever policy prescriptions that we

<|endoftext|> perspective . " What is

Decoder Weights Distribution

Head 0: 0.08

Head 1: 0.08

Head 2: 0.07

Head 3: 0.07

Head 4: 0.10

Head 5: 0.08

Head 6: 0.07

Head 7: 0.09

Head 8: 0.08

Head 9: 0.09

Head 10: 0.08

Head 11: 0.09

Positive logits

Conflict1.60

Scrolls1.47

Disclosure1.41

Illuminati1.34

Imperium1.33

Trance1.32

Bleach1.30

assic1.30

guiActive1.29

Cancel1.28

Noct1.24

"!1.24

Volume1.23

zbollah1.23

Flare1.23

Baal1.21

Isis1.21

Anonymous1.20

resa1.19

orget1.19

Negative logits

ertodd-1.56

lov-1.51

enegger-1.46

grounds-1.42

Recomm-1.41

stones-1.40

handedly-1.40

stairs-1.40

版-1.38

adj-1.38

eteria-1.37

ingham-1.37

stead-1.36

裏�-1.33

kson-1.32

ington-1.30

atively-1.30

ois-1.29

Cost-1.28

mable-1.27

INTERVAL 0.007 - 0.008
CONTAINS 0.000%

rou ng ing for money . He stayed on his corner

INTERVAL 0.006 - 0.007
CONTAINS 0.000%

INTERVAL 0.006 - 0.006
CONTAINS 0.000%

INTERVAL 0.005 - 0.006
CONTAINS 0.000%

INTERVAL 0.004 - 0.005
CONTAINS 0.000%

INTERVAL 0.003 - 0.004
CONTAINS 0.000%

INTERVAL 0.002 - 0.003
CONTAINS 0.000%

INTERVAL 0.002 - 0.002
CONTAINS 0.000%

INTERVAL 0.001 - 0.002
CONTAINS 0.000%

INTERVAL 0.000 - 0.001
CONTAINS 100.000%

Ministry of Education , told Al - Monitor the Arabic language

Ļ s budget architect , sets his sights on fighting poverty

p lex ig lass . For example , you have bars

it was purely a top - down game , right ?

Yet , the English - speaking world remains behold en to

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 25: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.951

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

. " What is true , though , is that whatever

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.08

Head 1: 0.09

Head 2: 0.08

Head 3: 0.07

Head 4: 0.08

Head 5: 0.10

Head 6: 0.08

Head 7: 0.07

Head 8: 0.07

Head 9: 0.07

Head 10: 0.08

Head 11: 0.13

Positive logits

territ1.85

opio1.60

ilers1.59

iler1.52

disparate1.52

ּ1.47

heterogeneity1.47

resid1.46

Sector1.45

Donetsk1.44

CPC1.44

habitable1.42

consolidation1.41

prevailing1.40

flattened1.40

looph1.39

locality1.39

millenn1.38

disl1.38

departures1.38

Negative logits

anus-1.87

aunt-1.74

helm-1.69

tube-1.59

abel-1.59

omorphic-1.57

appropriate-1.56

nant-1.56

eus-1.54

iven-1.53

atever-1.50

puff-1.49

grid-1.49

upload-1.49

rend-1.47

ife-1.47

xt-1.46

bourg-1.46

Puzz-1.45

your-1.45

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

over the years , women have also increasingly taken more steps

uers ' may never go into mine ' Ċ Ċ Updated

sector ( including more than $ 100 million cut from the

arrests occurred Thursday afternoon . Ċ Ċ Having made national news

core idea is <|endoftext|> By Gene Nelson , Ph . D

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 26: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.263

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

we 've been proposing don 't reach , are not heard

<|endoftext|> perspective . " What is

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

<|endoftext|> perspective . " What is

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

<|endoftext|> perspective . " What is

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.09

Head 1: 0.09

Head 2: 0.08

Head 3: 0.08

Head 4: 0.08

Head 5: 0.08

Head 6: 0.09

Head 7: 0.08

Head 8: 0.07

Head 9: 0.08

Head 10: 0.07

Head 11: 0.10

Positive logits

Intelligent1.62

Compass1.58

SAP1.47

Pric1.47

®1.45

iP1.45

Rail1.45

Disability1.44

LOW1.43

Econom1.42

Jaw1.38

smart1.36

Million1.36

Banking1.33

IoT1.32

marg1.32

Insurance1.31

Wealth1.31

Rising1.30

appliances1.29

Negative logits

ulkan-1.74

ּ-1.71

archive-1.66

ioch-1.64

Cheong-1.60

Ibid-1.57

rosso-1.56

fman-1.55

edit-1.54

andestine-1.50

isi-1.49

achu-1.48

interstitial-1.48

scrimmage-1.47

COURT-1.47

clave-1.46

ongyang-1.46

hent-1.45

igmatic-1.43

ortmund-1.43

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

use of " her p der p " or " der

, brand confusion is the number one brand killer . Pers

ancient Cel ts once did , thanks to a new analysis

of Air . Ċ Ċ Reward # 2 - N umbered

of folklore to traditions from different places that work in totally

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 27: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.347

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

<|endoftext|> perspective . " What is

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

<|endoftext|> perspective . " What is

true , though , is that whatever policy prescriptions that we

<|endoftext|> perspective . " What is

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.10

Head 1: 0.10

Head 2: 0.07

Head 3: 0.07

Head 4: 0.07

Head 5: 0.08

Head 6: 0.09

Head 7: 0.09

Head 8: 0.07

Head 9: 0.07

Head 10: 0.07

Head 11: 0.11

Positive logits

ldon1.67

Quinn1.61

resigned1.56

deen1.55

Dag1.50

foster1.48

intercepted1.43

delegated1.39

xual1.37

attend1.35

Dian1.35

baptized1.34

rented1.34

Seah1.33

Berm1.31

coached1.31

Eliot1.31

Bridgewater1.30

freelance1.29

verson1.29

Negative logits

Versions-1.79

SOURCE-1.69

PASS-1.64

GMT-1.64

IPM-1.64

Nit-1.64

Recipe-1.61

Ingredients-1.61

uration-1.57

Enlarge-1.56

Methods-1.55

Instead-1.55

Availability-1.54

MX-1.52

Exit-1.52

Appearance-1.52

Installation-1.51

ATIONS-1.51

TextColor-1.51

NOW-1.51

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

findings with Be ed ie âĢ Ļ s conclusion that only

over 400 , 000 people who were born in New Zealand

growth to come , even this year . He predicts a

that needs 300 + carries , or even 250 to be

pe ering out the window of one of the No Crack

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 28: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.439

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

<|endoftext|> perspective . " What is true , though ,

<|endoftext|> perspective . " What is true , though , is

<|endoftext|> perspective . " What is

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

<|endoftext|> perspective . " What is

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.10

Head 1: 0.09

Head 2: 0.07

Head 3: 0.07

Head 4: 0.08

Head 5: 0.08

Head 6: 0.10

Head 7: 0.08

Head 8: 0.09

Head 9: 0.07

Head 10: 0.06

Head 11: 0.10

Positive logits

Pink1.75

Pie1.75

Indigo1.66

Shir1.66

Defin1.62

Jou1.62

Carly1.59

‎1.55

Unch1.55

Bak1.54

Pale1.54

Pyr1.53

Sins1.50

Riy1.49

Beet1.49

Rae1.49

Nep1.48

Fried1.48

Ide1.47

Sa1.47

Negative logits

orius-1.82

handler-1.64

elong-1.56

finding-1.52

rotation-1.51

collided-1.50

emaker-1.48

otation-1.47

dummy-1.44

gee-1.43

tether-1.42

radius-1.41

farther-1.41

ality-1.41

imeter-1.38

consecut-1.37

decomp-1.37

rect-1.37

trial-1.36

otomy-1.35

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

and ( 2 ) the liberal view that the Voting Rights

have an event in " at least one city " in

. Ċ Ċ The E els were also reprim anded for

eeds owner Mass imo Cell ino reveals that C ag li

, thinking I 'd be safer there than walking home alone

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Feature 29: Dead

TOP ACTIVATIONS
MAX = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Top DFA by src position
MAX = 0.379

proposing don 't reach , are not heard by , the

<|endoftext|> perspective . " What is

we 've been proposing don 't reach , are not heard

<|endoftext|> perspective . " What is

that whatever policy prescriptions that we 've been proposing don 't

<|endoftext|> perspective . " What is

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

<|endoftext|> perspective . " What is

is true , though , is that whatever policy prescriptions that

Decoder Weights Distribution

Head 0: 0.08

Head 1: 0.09

Head 2: 0.09

Head 3: 0.08

Head 4: 0.07

Head 5: 0.09

Head 6: 0.10

Head 7: 0.08

Head 8: 0.08

Head 9: 0.08

Head 10: 0.08

Head 11: 0.09

Positive logits

��1.77

addictive1.55

��1.54

esty1.49

xual1.48

synergy1.46

staking1.44

flavour1.40

Duck1.38

hopped1.37

tomat1.36

itialized1.35

Sour1.35

oats1.35

broccoli1.32

enos1.31

ERO1.30

onut1.29

CHAT1.28

irez1.27

Negative logits

owe-1.60

marqu-1.52

onga-1.45

ocity-1.39

�-1.36

she-1.35

morrow-1.35

he-1.33

�-1.33

etheus-1.32

hum-1.31

minister-1.31

maid-1.30

roy-1.29

nm-1.28

sidew-1.25

ilings-1.25

etimes-1.24

ribution-1.23

amina-1.21

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

Rutgers coaching staff before selecting P isc at away as his

need new hardware to take advantage of them . You can

that way too often . Ċ Ċ We have done a

yet beautiful scream that cuts through the noise and stops the

was eager to remain involved in motor racing in some capacity

BOTTOM ACTIVATIONS
MIN = 0.000

proposing don 't reach , are not heard by , the

been proposing don 't reach , are not heard by ,

we 've been proposing don 't reach , are not heard

've been proposing don 't reach , are not heard by

perspective . " What is true , though , is that

<|endoftext|> perspective . " What is true , though , is

. " What is true , though , is that whatever

" What is true , though , is that whatever policy

that whatever policy prescriptions that we 've been proposing don 't

is that whatever policy prescriptions that we 've been proposing don

though , is that whatever policy prescriptions that we 've been

, is that whatever policy prescriptions that we 've been proposing

policy prescriptions that we 've been proposing don 't reach ,

whatever policy prescriptions that we 've been proposing don 't reach

prescriptions that we 've been proposing don 't reach , are

that we 've been proposing don 't reach , are not

, though , is that whatever policy prescriptions that we 've

true , though , is that whatever policy prescriptions that we

What is true , though , is that whatever policy prescriptions

is true , though , is that whatever policy prescriptions that

Token	`proposing`
Feature activation	+0.000

Token	`don`
Feature activation	+0.000

Token	`'t`
Feature activation	+0.000

Token	`reach`
Feature activation	+0.000

Token	`,`
Feature activation	+0.000

Feature 0: Dead

TOP ACTIVATIONSMAX = 0.000

Top DFA by src positionMAX = 0.462

Decoder Weights Distribution

Positive logits

Negative logits

INTERVAL 0.000 - 0.000CONTAINS 100.000%

BOTTOM ACTIVATIONSMIN = 0.000

Feature 1: Uninterpretable

TOP ACTIVATIONSMAX = 5.379

Top DFA by src positionMAX = 6.777

Decoder Weights Distribution

Positive logits

Negative logits

INTERVAL 4.841 - 5.379CONTAINS 0.000%

INTERVAL 4.304 - 4.841CONTAINS 0.000%

INTERVAL 3.766 - 4.304CONTAINS 0.000%

INTERVAL 3.228 - 3.766CONTAINS 0.000%

INTERVAL 2.690 - 3.228CONTAINS 0.000%

INTERVAL 2.152 - 2.690CONTAINS 0.000%

INTERVAL 1.614 - 2.152CONTAINS 0.000%

INTERVAL 1.076 - 1.614CONTAINS 0.000%

INTERVAL 0.538 - 1.076CONTAINS 0.000%

INTERVAL 0.000 - 0.538CONTAINS 99.999%

BOTTOM ACTIVATIONSMIN = 0.000

Feature 2: In word / title starting with " Som"

TOP ACTIVATIONSMAX = 3.276

Top DFA by src positionMAX = 5.412

Decoder Weights Distribution

Positive logits

Negative logits

INTERVAL 2.949 - 3.276CONTAINS 0.000%

INTERVAL 2.621 - 2.949CONTAINS 0.000%

INTERVAL 2.293 - 2.621CONTAINS 0.000%

INTERVAL 1.966 - 2.293CONTAINS 0.000%

INTERVAL 1.638 - 1.966CONTAINS 0.000%

INTERVAL 1.311 - 1.638CONTAINS 0.000%

INTERVAL 0.983 - 1.311CONTAINS 0.000%

INTERVAL 0.655 - 0.983CONTAINS 0.000%

INTERVAL 0.328 - 0.655CONTAINS 0.000%

INTERVAL 0.000 - 0.328CONTAINS 99.998%

BOTTOM ACTIVATIONSMIN = 0.000

Feature 3: Follows "Def"

TOP ACTIVATIONSMAX = 4.369

Top DFA by src positionMAX = 5.434

Decoder Weights Distribution

Positive logits

Negative logits

INTERVAL 3.932 - 4.369CONTAINS 0.000%

INTERVAL 3.495 - 3.932CONTAINS 0.000%

INTERVAL 3.058 - 3.495CONTAINS 0.000%

INTERVAL 2.621 - 3.058CONTAINS 0.000%

INTERVAL 2.184 - 2.621CONTAINS 0.000%

INTERVAL 1.747 - 2.184CONTAINS 0.000%

INTERVAL 1.311 - 1.747CONTAINS 0.000%

INTERVAL 0.874 - 1.311CONTAINS 0.000%

INTERVAL 0.437 - 0.874CONTAINS 0.001%

INTERVAL 0.000 - 0.437CONTAINS 99.998%

BOTTOM ACTIVATIONSMIN = 0.000

Feature 4: Dead

TOP ACTIVATIONSMAX = 0.000

Top DFA by src positionMAX = 0.334

Decoder Weights Distribution

Positive logits

Negative logits

INTERVAL 0.000 - 0.000CONTAINS 100.000%

BOTTOM ACTIVATIONSMIN = 0.000

Feature 5: Dead

TOP ACTIVATIONSMAX = 0.000

Top DFA by src positionMAX = 0.225

Decoder Weights Distribution

Positive logits

Negative logits

INTERVAL 0.000 - 0.000CONTAINS 100.000%

BOTTOM ACTIVATIONSMIN = 0.000

Feature 6: " human" token

TOP ACTIVATIONSMAX = 1.749

Top DFA by src positionMAX = 3.567

Decoder Weights Distribution

Positive logits

TOP ACTIVATIONS
MAX = 0.000

Top DFA by src position
MAX = 0.462

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 5.379

Top DFA by src position
MAX = 6.777

INTERVAL 4.841 - 5.379
CONTAINS 0.000%

INTERVAL 4.304 - 4.841
CONTAINS 0.000%

INTERVAL 3.766 - 4.304
CONTAINS 0.000%

INTERVAL 3.228 - 3.766
CONTAINS 0.000%

INTERVAL 2.690 - 3.228
CONTAINS 0.000%

INTERVAL 2.152 - 2.690
CONTAINS 0.000%

INTERVAL 1.614 - 2.152
CONTAINS 0.000%

INTERVAL 1.076 - 1.614
CONTAINS 0.000%

INTERVAL 0.538 - 1.076
CONTAINS 0.000%

INTERVAL 0.000 - 0.538
CONTAINS 99.999%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 3.276

Top DFA by src position
MAX = 5.412

INTERVAL 2.949 - 3.276
CONTAINS 0.000%

INTERVAL 2.621 - 2.949
CONTAINS 0.000%

INTERVAL 2.293 - 2.621
CONTAINS 0.000%

INTERVAL 1.966 - 2.293
CONTAINS 0.000%

INTERVAL 1.638 - 1.966
CONTAINS 0.000%

INTERVAL 1.311 - 1.638
CONTAINS 0.000%

INTERVAL 0.983 - 1.311
CONTAINS 0.000%

INTERVAL 0.655 - 0.983
CONTAINS 0.000%

INTERVAL 0.328 - 0.655
CONTAINS 0.000%

INTERVAL 0.000 - 0.328
CONTAINS 99.998%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 4.369

Top DFA by src position
MAX = 5.434

INTERVAL 3.932 - 4.369
CONTAINS 0.000%

INTERVAL 3.495 - 3.932
CONTAINS 0.000%

INTERVAL 3.058 - 3.495
CONTAINS 0.000%

INTERVAL 2.621 - 3.058
CONTAINS 0.000%

INTERVAL 2.184 - 2.621
CONTAINS 0.000%

INTERVAL 1.747 - 2.184
CONTAINS 0.000%

INTERVAL 1.311 - 1.747
CONTAINS 0.000%

INTERVAL 0.874 - 1.311
CONTAINS 0.000%

INTERVAL 0.437 - 0.874
CONTAINS 0.001%

INTERVAL 0.000 - 0.437
CONTAINS 99.998%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 0.000

Top DFA by src position
MAX = 0.334

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 0.000

Top DFA by src position
MAX = 0.225

INTERVAL 0.000 - 0.000
CONTAINS 100.000%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 1.749

Top DFA by src position
MAX = 3.567

INTERVAL 1.574 - 1.749
CONTAINS 0.000%

INTERVAL 1.399 - 1.574
CONTAINS 0.000%

INTERVAL 1.224 - 1.399
CONTAINS 0.002%

INTERVAL 1.049 - 1.224
CONTAINS 0.005%

INTERVAL 0.874 - 1.049
CONTAINS 0.006%

INTERVAL 0.700 - 0.874
CONTAINS 0.004%

INTERVAL 0.525 - 0.700
CONTAINS 0.003%

INTERVAL 0.350 - 0.525
CONTAINS 0.002%

INTERVAL 0.175 - 0.350
CONTAINS 0.002%

INTERVAL 0.000 - 0.175
CONTAINS 99.977%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 1.372

Top DFA by src position
MAX = 2.696

INTERVAL 1.235 - 1.372
CONTAINS 0.000%

INTERVAL 1.097 - 1.235
CONTAINS 0.000%

INTERVAL 0.960 - 1.097
CONTAINS 0.000%

INTERVAL 0.823 - 0.960
CONTAINS 0.000%

INTERVAL 0.686 - 0.823
CONTAINS 0.000%

INTERVAL 0.549 - 0.686
CONTAINS 0.001%

INTERVAL 0.412 - 0.549
CONTAINS 0.001%

INTERVAL 0.274 - 0.412
CONTAINS 0.001%

INTERVAL 0.137 - 0.274
CONTAINS 0.000%

INTERVAL 0.000 - 0.137
CONTAINS 99.997%

BOTTOM ACTIVATIONS
MIN = 0.000

TOP ACTIVATIONS
MAX = 0.000

Top DFA by src position
MAX = 0.638

INTERVAL 0.000 - 0.000
CONTAINS 100.000%