INDEX

Explanations

phrases containing words associated with offense or insensitivity

oai_token-act-pair · gpt-3.5-turbo

language related to offensiveness or inappropriate content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

uilding

-0.81

ills

-0.75

arten

-0.75

negie

-0.66

frames

-0.65

runner

-0.64

knit

-0.64

inding

-0.64

ail

-0.64

enfranch

-0.64

POSITIVE LOGITS

 insult

1.03

 remarks

1.00

 slurs

1.00

 insults

0.93

 caric

0.89

 humor

0.89

 disrespect

0.88

 humour

0.88

 stereotypes

0.88

 insulting

0.86

Activations Density 0.134%

phrases containing words associated with offense or insensitivity

language related to offensiveness or inappropriate content

No Comments

No Known Activations