INDEX

Explanations

words related to inappropriate or offensive content

oai_token-act-pair · gpt-3.5-turbo

terms related to indecency and explicit content

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

rod

-0.87

rule

-0.77

Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯

-0.76

abil

-0.75

Ability

-0.75

iller

-0.75

OE

-0.73

rador

-0.73

ONES

-0.71

Reviewed

-0.71

POSITIVE LOGITS

 lewd

1.12

 indecent

1.11

 nudity

1.04

 nude

1.03

uously

0.88

 genital

0.81

 genitals

0.79

 exhib

0.78

 scenes

0.78

 underwear

0.78

Activations Density 0.022%

words related to inappropriate or offensive content

terms related to indecency and explicit content

No Comments

No Known Activations