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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 specialization

-0.73

coe

-0.65

è¡

-0.63

mble

-0.62

 indebted

-0.62

ĸļ

-0.62

 unaff

-0.61

çĲ

-0.60

 organisation

-0.59

æĺ

-0.58

POSITIVE LOGITS

eat

0.75

emi

0.74

empt

0.73

vision

0.72

ItemThumbnailImage

0.71

 Olson

0.70

hap

0.70

agically

0.68

eries

0.68

abl

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.