INDEX

Explanations

references to political or violent acts, specifically related to assassination

oai_token-act-pair · gpt-3.5-turbo

references to assassinations and assassination attempts

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Cola

-0.83

Va

-0.77

TAG

-0.77

STRUCT

-0.73

Vert

-0.73

HUD

-0.71

bos

-0.70

OPER

-0.70

)=(

-0.69

comments

-0.68

POSITIVE LOGITS

 assass

1.29

 assassination

1.28

 assassin

1.26

 assassinated

1.24

 assassins

1.21

 Assass

1.08

 assassinate

1.06

 mosqu

0.89

inations

0.84

inating

0.84

Activations Density 0.016%

references to political or violent acts, specifically related to assassination

references to assassinations and assassination attempts

No Comments

No Known Activations