INDEX

Explanations

mentions of gender-related words and topics

oai_token-act-pair · gpt-3.5-turbo

references to gender

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Alert

-0.76

 Bullets

-0.72

ateurs

-0.72

aching

-0.70

thus

-0.68

 Alert

-0.68

rich

-0.68

Burn

-0.68

ached

-0.67

runs

-0.67

POSITIVE LOGITS

 gender

3.59

 Gender

3.01

Gender

2.91

gender

2.79

 genders

2.78

 sexes

2.04

 sexism

1.96

 femin

1.94

sex

1.90

 transgender

1.83

Activations Density 0.015%

mentions of gender-related words and topics

references to gender

No Comments

No Known Activations