INDEX

Explanations

names or terms related to public figures

oai_token-act-pair · gpt-3.5-turbo

names or words that end with the suffix 'ette'

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

akening

-0.77

ific

-0.71

iasm

-0.70

reddits

-0.70

ority

-0.66

alez

-0.65

ĻĤ

-0.63

raphics

-0.63

 mathemat

-0.62

ained

-0.62

POSITIVE LOGITS

ette

1.59

ettes

1.39

ttes

0.89

lla

0.87

 Scotia

0.85

Mania

0.84

brate

0.83

brates

0.78

ery

0.78

 Hebdo

0.76

Activations Density 0.008%

names or terms related to public figures

names or words that end with the suffix 'ette'

No Comments

No Known Activations