INDEX

Explanations

the word "potentially" followed by a description of a possible risk, threat, or consequence

oai_token-act-pair · gpt-3.5-turbo

references to potential risks or threats

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ger

-0.85

board

-0.79

baugh

-0.75

downs

-0.72

boards

-0.71

geist

-0.71

gers

-0.71

rike

-0.69

 Gore

-0.67

io

-0.67

POSITIVE LOGITS

 feas

0.95

 exting

0.93

 vulner

0.89

 conclud

0.89

 metic

0.87

 jeopard

0.87

 unbeliev

0.85

 exha

0.83

 notor

0.83

 susceptible

0.83

Activations Density 0.006%

the word "potentially" followed by a description of a possible risk, threat, or consequence

references to potential risks or threats

No Comments

No Known Activations