INDEX

Explanations

phrases related to violent or harmful actions

oai_token-act-pair · gpt-3.5-turbo

references to violence and atrocities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ellar

-0.84

cius

-0.80

 facult

-0.78

paren

-0.78

isphere

-0.71

ria

-0.71

oard

-0.69

 workspace

-0.69

 latitude

-0.68

orers

-0.68

POSITIVE LOGITS

 inflicted

1.00

 havoc

0.88

 Slaughter

0.80

 devastation

0.78

 rampage

0.78

 blows

0.78

 spree

0.76

à¨

0.74

Mum

0.71

soever

0.70

Activations Density 0.149%

phrases related to violent or harmful actions

references to violence and atrocities

No Comments

No Known Activations