INDEX

Explanations

profane and derogatory terms

oai_token-act-pair · gpt-3.5-turbo

derogatory terms and phrases related to unflattering behaviors or characteristics

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

 Corm

-0.64

Source

-0.64

WER

-0.63

 Prophe

-0.61

 Novel

-0.61

 divisions

-0.59

ãģ®ç

-0.59

Feature

-0.59

rics

-0.57

 Defender

-0.57

POSITIVE LOGITS

jer

1.17

 jerk

1.14

usalem

1.03

weed

0.86

offs

0.82

boa

0.82

ometer

0.80

ety

0.80

bucks

0.80

itude

0.79

Activations Density 0.021%

profane and derogatory terms

derogatory terms and phrases related to unflattering behaviors or characteristics

No Comments

No Known Activations