INDEX

Explanations

people's names or positions in political settings

oai_token-act-pair · gpt-3.5-turbo

references to media outlets and news reporting entities

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

Not in Any Lists

No Comments

Negative Logits

gone

-0.84

ãĤ§

-0.70

gran

-0.65

ĪĴ

-0.65

ãĥĺ

-0.65

ãĥķãĤ¡

-0.63

cius

-0.63

gnu

-0.63

chest

-0.62

ãĤ¡

-0.62

POSITIVE LOGITS

 Photo

1.03

 PHOTO

1.00

 Images

0.97

 IMAGES

0.97

 VIDEOS

0.95

Buy

0.81

 Caption

0.81

photo

0.80

Image

0.78

 FILE

0.77

Activations Density 0.035%

people's names or positions in political settings

references to media outlets and news reporting entities

No Comments

No Known Activations