INDEX

Explanations

verbs related to taking back or withdrawing something

oai_token-act-pair · gpt-3.5-turbo

terms related to retraction or correction of published content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

amac

-0.89

pn

-0.84

places

-0.82

atown

-0.78

swick

-0.77

arding

-0.77

eers

-0.76

drivers

-0.73

fire

-0.73

eson

-0.71

POSITIVE LOGITS

 retract

0.98

raction

0.87

ractions

0.86

 retracted

0.83

 guiActive

0.71

 unsub

0.70

ŃĶ

0.69

 confessions

0.69

iple

0.68

 facult

0.66

Activations Density 0.020%

verbs related to taking back or withdrawing something

terms related to retraction or correction of published content

No Comments

No Known Activations