INDEX

Explanations

negative or offensive language and comments

oai_token-act-pair · gpt-3.5-turbo

references to offensive comments and remarks

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

oglu

-0.74

negie

-0.73

prus

-0.72

sonian

-0.71

Luck

-0.69

UNCH

-0.69

iets

-0.69

inav

-0.69

aer

-0.68

 Luck

-0.68

POSITIVE LOGITS

 inappropriate

1.13

 inappropriately

1.13

 slurs

1.09

 lewd

1.05

 disrespectful

1.04

 indecent

1.02

 abusive

1.01

 harassing

1.00

 misogyn

0.95

 uttered

0.94

Activations Density 0.349%

negative or offensive language and comments

references to offensive comments and remarks

No Comments

No Known Activations