INDEX

Explanations

phrases related to photographs or images

oai_token-act-pair · gpt-3.5-turbo

instances of the word "pictured" or its variations in contexts related to images or photographs

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

cession

-0.86

cknow

-0.78

ced

-0.78

die

-0.78

angan

-0.71

vernment

-0.68

ance

-0.68

ammy

-0.67

statement

-0.65

gettable

-0.64

POSITIVE LOGITS

 pictured

1.15

 photographed

0.88

eus

0.82

 Pict

0.79

 Gallery

0.78

 IMAGES

0.77

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

0.75

lihood

0.75

ãĤ´

0.72

sburg

0.71

Activations Density 0.005%

phrases related to photographs or images

instances of the word "pictured" or its variations in contexts related to images or photographs

No Comments

No Known Activations