INDEX

Explanations

photographs or images as described by the activation of the word "Photo"

oai_token-act-pair · gpt-3.5-turbo

instances of image or photo references in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

uckle

-0.82

powers

-0.79

anches

-0.77

ims

-0.75

anche

-0.74

resses

-0.74

apers

-0.73

astically

-0.72

uchs

-0.72

merce

-0.71

POSITIVE LOGITS

 Photo

1.03

Photo

0.96

 Album

0.90

journal

0.90

 Thumbnails

0.89

 illustration

0.89

 Courtesy

0.89

 Photographer

0.87

 Editing

0.86

 Photograph

0.86

Activations Density 0.017%

photographs or images as described by the activation of the word "Photo"

instances of image or photo references in the text

No Comments

No Known Activations