INDEX

Explanations

the word "man" with different levels of activation, focusing on descriptions or actions related to a man

oai_token-act-pair · gpt-3.5-turbo

mentions of the word "man."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

soType

-0.83

ettings

-0.67

emies

-0.66

agall

-0.63

ances

-0.63

 Cosponsors

-0.62

iances

-0.62

ancing

-0.61

alysis

-0.61

 Lerner

-0.61

POSITIVE LOGITS

man

3.40

 woman

1.98

Man

1.91

men

1.90

guy

1.80

Man

1.78

 mans

1.72

boy

1.72

 dude

1.69

man

1.66

Activations Density 0.042%

the word "man" with different levels of activation, focusing on descriptions or actions related to a man

mentions of the word "man."

No Comments

No Known Activations