INDEX

Explanations

vulgar and offensive terms

oai_token-act-pair · gpt-3.5-turbo

references to sexual or vulgar terms

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

iation

-0.83

othy

-0.75

ril

-0.75

aneously

-0.75

GRE

-0.74

VERTISEMENT

-0.73

ORGE

-0.68

reek

-0.67

OR

-0.66

oS

-0.66

POSITIVE LOGITS

ussy

0.97

ignt

0.82

 panties

0.80

holes

0.79

cat

0.79

hole

0.79

 Riot

0.78

essee

0.77

chet

0.77

 lips

0.76

Activations Density 0.011%

vulgar and offensive terms

references to sexual or vulgar terms

No Comments

No Known Activations