INDEX

Explanations

phrases or words related to harmfulness

oai_token-act-pair · gpt-3.5-turbo

references to harmful effects or substances

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

vet

-0.81

azon

-0.69

Square

-0.65

Option

-0.65

Finish

-0.64

antry

-0.63

 finally

-0.62

istan

-0.60

gin

-0.60

 Option

-0.59

POSITIVE LOGITS

 harmful

3.33

 detrimental

2.30

 damaging

2.03

 harms

1.88

 destructive

1.87

 harming

1.82

 dangerous

1.78

 poisonous

1.76

 beneficial

1.68

 hazardous

1.67

Activations Density 0.020%

phrases or words related to harmfulness

references to harmful effects or substances

No Comments

No Known Activations