INDEX

Explanations

content or objects that are physically covered or obscured in some way

oai_token-act-pair · gpt-3.5-turbo

instances of the word "covered" in various contexts

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

STEM

-0.65

 RELE

-0.62

 fortune

-0.62

ELL

-0.60

ASY

-0.60

Hunt

-0.58

----------

-0.58

å§

-0.58

cial

-0.58

 pleasing

-0.57

POSITIVE LOGITS

ummer

0.78

 cover

0.76

Cover

0.76

usky

0.74

alls

0.73

fold

0.73

 Coverage

0.73

toe

0.72

 covers

0.71

 extensively

0.70

Activations Density 0.037%

content or objects that are physically covered or obscured in some way

instances of the word "covered" in various contexts

No Comments

No Known Activations