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

 censorship

-0.69

 spam

-0.69

xx

-0.65

mob

-0.64

ffect

-0.64

 backdoor

-0.63

 interven

-0.62

oldown

-0.61

ction

-0.60

llah

-0.60

POSITIVE LOGITS

Atl

0.93

 Parables

0.93

 Palest

0.90

angan

0.72

eport

0.71

FontSize

0.70

 Orth

0.69

 Catalan

0.66

vec

0.66

beit

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.