INDEX

Explanations

information about articles or sections labeled "About."

oai_token-act-pair · gpt-3.5-turbo

the word "About" indicating introductory phrases or sections

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 jugg

-0.70

 shock

-0.70

 leap

-0.68

 breath

-0.64

tit

-0.63

dod

-0.62

 shadow

-0.61

 litter

-0.61

red

-0.61

 control

-0.61

POSITIVE LOGITS

About

3.74

 About

2.08

about

1.91

 ABOUT

1.87

Around

1.58

abouts

1.31

Overview

1.19

 Approximately

1.19

Nearly

1.18

Details

1.14

Activations Density 0.018%

information about articles or sections labeled "About."

the word "About" indicating introductory phrases or sections

No Comments

No Known Activations