INDEX

Explanations

No Explanations Found

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

 buyer

-0.66

 Ceres

-0.66

 Pebble

-0.66

catentry

-0.65

 wallpaper

-0.64

 Password

-0.64

ebus

-0.63

 ISBN

-0.62

 vulnerability

-0.61

 captcha

-0.61

POSITIVE LOGITS

tyard

0.92

 Ames

0.79

ais

0.77

ersen

0.76

forts

0.72

arer

0.72

iers

0.71

©¶æ

0.71

Ã¢

0.69

ank

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.