INDEX

Explanations

terms related to LGBTQ+ identities and advocacy

oai_token-act-pair · gpt-3.5-turbo

references to LGBTQ+ identities and issues

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

 snowball

-0.56

 viewer

-0.55

 Mous

-0.54

Redditor

-0.51

 Unch

-0.51

 spur

-0.50

witz

-0.49

é¾į

-0.48

 tightened

-0.48

pload

-0.48

POSITIVE LOGITS

etc

1.28

etc

1.11

âĢ¦)

0.84

âĢ¦

0.81

 âĢ¦

0.79

ect

0.77

âĢ¦.

0.76

...)

0.73

0.70

 welf

0.70

Activations Density 0.179%

terms related to LGBTQ+ identities and advocacy

references to LGBTQ+ identities and issues

No Comments

No Known Activations