INDEX

Explanations

words related to violent actions like assassination, hacking, bombing, and killing

oai_token-act-pair · gpt-3.5-turbo

references to acts of violence and targeted attacks

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

-0.84

izont

-0.78

itte

-0.78

mie

-0.76

arine

-0.75

 Moder

-0.73

ellar

-0.72

BuyableInstoreAndOnline

-0.69

ertain

-0.69

ube

-0.69

POSITIVE LOGITS

 spree

1.40

 rampage

1.06

 allegation

1.00

 scandal

0.99

 attempt

0.99

 frenzy

0.97

 accusation

0.95

 affair

0.91

 massacre

0.88

of

0.87

Activations Density 0.224%

words related to violent actions like assassination, hacking, bombing, and killing

references to acts of violence and targeted attacks

No Comments

No Known Activations