INDEX

Explanations

No Explanations Found

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

externalActionCode

-0.77

Registered

-0.63

nee

-0.62

lihood

-0.62

 Confederation

-0.60

 unwitting

-0.60

uable

-0.60

erous

-0.59

Prime

-0.59

Demand

-0.58

POSITIVE LOGITS

 interests

0.86

 arsen

0.79

enhagen

0.72

azes

0.71

 arcs

0.67

 boundaries

0.66

rients

0.66

 Traps

0.66

cha

0.66

bane

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.