INDEX

Explanations

phrases related to destructive or criminal activities

oai_token-act-pair · gpt-3.5-turbo

terminology related to destruction and damage in various contexts

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

•GPT2 L10 resid: 'LRRRRRLLRRRL'[:-3]

No Comments

Negative Logits

iasis

-0.72

ois

-0.72

algia

-0.69

idth

-0.68

METHOD

-0.66

 sunshine

-0.66

 renaissance

-0.65

reprene

-0.64

ITNESS

-0.64

oneliness

-0.63

POSITIVE LOGITS

 causing

0.98

 belonging

0.97

 guarding

0.88

 unlawfully

0.84

 injuring

0.84

 indiscrim

0.80

 prematurely

0.79

 storing

0.78

 destroying

0.76

 violently

0.76

Activations Density 0.348%

phrases related to destructive or criminal activities

terminology related to destruction and damage in various contexts

No Comments

No Known Activations