INDEX

Explanations

the word "topic" at varying strengths

oai_token-act-pair · gpt-3.5-turbo

references to specific subjects or themes being discussed

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

ignt

-0.73

 Yates

-0.71

leeve

-0.71

 Zimmer

-0.70

igned

-0.69

ramid

-0.66

zik

-0.65

inyl

-0.63

 Towers

-0.63

xon

-0.63

POSITIVE LOGITS

 topics

0.93

topic

0.89

Topic

0.88

matter

0.86

 topic

0.86

TextColor

0.85

 Topics

0.82

icular

0.78

bub

0.78

Topics

0.78

Activations Density 0.015%

the word "topic" at varying strengths

references to specific subjects or themes being discussed

No Comments

No Known Activations