INDEX

Explanations

adjectives or phrases describing potential harm or risks

oai_token-act-pair · gpt-3.5-turbo

references to danger and potentially harmful situations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

olitan

-0.82

roma

-0.79

ļéĨĴ

-0.78

mination

-0.76

via

-0.75

oration

-0.75

zzo

-0.75

hew

-0.74

arthed

-0.74

cedented

-0.73

POSITIVE LOGITS

 adolesc

0.91

 endanger

0.81

 undermin

0.75

 sounding

0.74

nesses

0.73

 dangerous

0.73

 threats

0.71

 overdose

0.71

 combination

0.70

 Danger

0.69

Activations Density 0.032%

adjectives or phrases describing potential harm or risks

references to danger and potentially harmful situations

No Comments

No Known Activations