INDEX

Explanations

websites, news agencies, and official entities mentioned in a text

oai_token-act-pair · gpt-3.5-turbo

references to various organizations and their reports

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

DragonMagazine

-0.65

ardless

-0.60

ictional

-0.59

 parity

-0.59

animate

-0.59

agame

-0.59

 contamin

-0.59

 disrupting

-0.57

anol

-0.57

aples

-0.56

POSITIVE LOGITS

 said

1.12

 wrote

1.03

said

1.01

 says

0.98

 told

0.95

 tells

0.91

 exclaimed

0.90

wrote

0.87

 quoted

0.86

Said

0.86

Activations Density 0.339%

websites, news agencies, and official entities mentioned in a text

references to various organizations and their reports

No Comments

No Known Activations