INDEX

Explanations

visual content or images in the text

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

images or captions in news articles

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

merce

-0.85

sole

-0.84

lies

-0.77

uckle

-0.75

ighter

-0.75

achus

-0.74

unts

-0.74

kees

-0.73

azes

-0.73

ugs

-0.73

POSITIVE LOGITS

 caption

1.04

Image

1.00

 Images

0.97

 Gallery

0.95

 Thumbnails

0.93

Images

0.90

 Image

0.90

 IMAGES

0.89

gallery

0.86

 Courtesy

0.83

Activations Density 0.013%

visual content or images in the text

images or captions in news articles

No Comments

No Known Activations