INDEX

Explanations

connections or ties to specific entities or groups, often with a negative connotation

oai_token-act-pair · gpt-3.5-turbo

connections to controversial or nefarious groups and individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

bra

-0.78

 partName

-0.76

meal

-0.74

ipeg

-0.73

facing

-0.70

erion

-0.70

ItemImage

-0.69

reci

-0.69

hog

-0.69

iculty

-0.68

POSITIVE LOGITS

 extremist

1.00

 billionaire

0.89

 extremism

0.84

 militant

0.82

 extremists

0.82

 Kremlin

0.80

 Osama

0.80

 wealthy

0.80

 terrorist

0.79

 discredited

0.79

Activations Density 0.267%

connections or ties to specific entities or groups, often with a negative connotation

connections to controversial or nefarious groups and individuals

No Comments

No Known Activations