INDEX

Explanations

threatening language or mentions of threats

oai_token-act-pair · gpt-3.5-turbo

mentions of threats to safety or security

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

igs

-0.72

 Cups

-0.70

OME

-0.70

neys

-0.68

neau

-0.67

 Stores

-0.66

 Gins

-0.65

 å¤

-0.64

abee

-0.63

abeth

-0.63

POSITIVE LOGITS

 threat

3.75

 threats

2.86

threat

2.85

 Threat

2.52

 menace

2.35

 danger

1.90

 threatening

1.80

 threaten

1.79

 threatened

1.67

 risk

1.44

Activations Density 0.018%

threatening language or mentions of threats

mentions of threats to safety or security

No Comments

No Known Activations