INDEX

Explanations

phrases related to harmful actions or situations

oai_token-act-pair · gpt-3.5-turbo

terms related to endangerment and harm to individuals or the public

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 negotiator

-0.68

 metab

-0.66

rera

-0.64

 grou

-0.63

 Explorer

-0.62

 Unch

-0.61

 contender

-0.61

 chamber

-0.59

 delinquent

-0.59

 backlog

-0.58

POSITIVE LOGITS

ings

0.98

ments

0.95

liest

0.91

ength

0.90

ences

0.85

ingly

0.79

ning

0.79

liness

0.79

lihood

0.78

ãĥĨ

0.78

Activations Density 0.032%

phrases related to harmful actions or situations

terms related to endangerment and harm to individuals or the public

No Comments

No Known Activations