INDEX

Explanations

strings of capitalized words, likely proper nouns such as names of places or people

oai_token-act-pair · gpt-3.5-turbo

common abbreviations or acronyms used in news or reports

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

stood

-0.71

 Abyssal

-0.64

 Haku

-0.63

doors

-0.63

except

-0.62

sed

-0.62

 Cerberus

-0.61

 Andersen

-0.61

 Schwar

-0.60

Tokens

-0.60

POSITIVE LOGITS

FORE

0.99

ONY

0.94

CLAIM

0.94

 SHARES

0.92

FER

0.90

HAM

0.90

VER

0.90

ENN

0.89

COL

0.89

BRE

0.88

Activations Density 0.089%

strings of capitalized words, likely proper nouns such as names of places or people

common abbreviations or acronyms used in news or reports

No Comments

No Known Activations