INDEX

Explanations

news headlines or article sections with sensational or alarming content

oai_token-act-pair · gpt-3.5-turbo

instances of the word "READ" or related calls to action

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

angers

-0.70

perties

-0.69

aea

-0.65

asketball

-0.64

phies

-0.63

 reconnect

-0.61

phia

-0.61

wine

-0.60

 Johns

-0.60

paid

-0.59

POSITIVE LOGITS

ING

1.27

INGS

1.22

 ALSO

1.15

TY

1.14

 MORE

1.14

BOOK

1.09

TING

1.09

NESS

1.07

ME

1.06

WOR

1.06

Activations Density 0.023%

news headlines or article sections with sensational or alarming content

instances of the word "READ" or related calls to action

No Comments

No Known Activations