INDEX

Explanations

phrases expressing a preference or support for something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "favor" in various contexts

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

wyn

-0.68

sea

-0.65

colour

-0.63

heid

-0.63

 Thames

-0.61

Australian

-0.60

yrs

-0.59

RAF

-0.59

anse

-0.59

NRL

-0.59

POSITIVE LOGITS

 favor

4.54

 favour

3.73

 favors

2.71

 Favor

2.31

 favoring

2.03

 favored

1.73

fav

1.72

 favorably

1.57

avor

1.46

 favorable

1.41

Activations Density 0.015%

phrases expressing a preference or support for something

instances of the word "favor" in various contexts

No Comments

No Known Activations