INDEX

Explanations

mentions of censorship or disapproval

oai_token-act-pair · gpt-3.5-turbo

references to censorship

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Dragonbound

-0.79

Goo

-0.79

Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯

-0.76

 Duchess

-0.73

 Donetsk

-0.71

 Vader

-0.71

ESA

-0.70

 Anthem

-0.70

 Rebellion

-0.69

 Finn

-0.69

POSITIVE LOGITS

 cens

1.16

orious

1.12

oring

1.07

orable

0.96

pling

0.92

urable

0.89

»

0.88

ĳ

0.88

orial

0.87

ile

0.86

Activations Density 0.011%

mentions of censorship or disapproval

references to censorship

No Comments

No Known Activations