INDEX

Explanations

mentions of the word "no" followed by various phrases

oai_token-act-pair · gpt-3.5-turbo

negation or denial phrases

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 nonetheless

-0.68

minster

-0.63

ially

-0.63

lus

-0.59

 Cathy

-0.58

 nevertheless

-0.58

ATIVE

-0.57

iership

-0.57

turned

-0.56

RED

-0.56

POSITIVE LOGITS

vel

0.99

zzle

0.93

otrop

0.86

obs

0.83

xious

0.83

 longer

0.81

vell

0.81

except

0.80

 warranties

0.79

isy

0.79

Activations Density 0.055%

mentions of the word "no" followed by various phrases

negation or denial phrases

No Comments

No Known Activations