INDEX

Explanations

words related to conflict, war, and violence

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "wr" and the word "scar."

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

uana

-0.69

 rapport

-0.68

ocial

-0.64

 assurances

-0.63

iment

-0.62

 automated

-0.62

 decriminal

-0.61

REDACTED

-0.61

 matured

-0.60

 openly

-0.60

POSITIVE LOGITS

wr

3.83

Wr

2.32

Wr

1.86

scar

1.42

WR

1.02

stru

1.00

ighton

0.99

 Wrath

0.95

Je

0.92

yk

0.92

Activations Density 0.064%

words related to conflict, war, and violence

instances of the prefix "wr" and the word "scar."

No Comments

No Known Activations