INDEX

Explanations

words related to someone being shamed, humiliated, or having a negative reputation

oai_token-act-pair · gpt-3.5-turbo

words related to disgrace and job loss

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

aminer

-0.79

itionally

-0.72

fw

-0.72

bra

-0.71

akeru

-0.71

vor

-0.71

ingham

-0.70

argo

-0.69

utra

-0.69

prints

-0.69

POSITIVE LOGITS

 embattled

1.19

 sacked

1.12

 disgr

1.06

 ousted

0.97

 embroiled

0.88

 estranged

0.88

 former

0.87

 thous

0.84

Former

0.82

 Former

0.78

Activations Density 0.016%

words related to someone being shamed, humiliated, or having a negative reputation

words related to disgrace and job loss

No Comments

No Known Activations