INDEX

Explanations

phrases related to violence or threats of harm

oai_token-act-pair · gpt-3.5-turbo

references to violence or death

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

angular

-0.79

herty

-0.78

 cancell

-0.77

RGB

-0.76

ARB

-0.75

taboola

-0.72

Applic

-0.71

ional

-0.70

soDeliveryDate

-0.70

Management

-0.68

POSITIVE LOGITS

 senseless

1.01

 aven

0.96

 vengeance

0.95

 revenge

0.94

 innocent

0.92

 martyr

0.87

 retribution

0.84

 innoc

0.83

 ransom

0.81

 coward

0.80

Activations Density 0.588%

phrases related to violence or threats of harm

references to violence or death

No Comments

No Known Activations