INDEX

Explanations

words associated with threats or potential danger

oai_token-act-pair · gpt-3.5-turbo

instances of the word "threaten"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Ded

-0.78

 graduating

-0.66

Fun

-0.65

omen

-0.64

 dedicated

-0.64

 Illum

-0.63

 illust

-0.63

 selected

-0.63

 commemor

-0.62

 retrospective

-0.62

POSITIVE LOGITS

 threaten

3.48

 threatens

2.62

threat

1.92

 threatening

1.76

 threatened

1.76

 threats

1.72

 menace

1.63

 endanger

1.59

 threat

1.55

 intimidate

1.48

Activations Density 0.015%

words associated with threats or potential danger

instances of the word "threaten"

No Comments

No Known Activations