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

 Banner

-0.79

 respondent

-0.73

istle

-0.67

 Huntington

-0.66

 Respond

-0.65

 Garrison

-0.65

 Participants

-0.63

 proble

-0.61

 Nass

-0.61

 Organizations

-0.60

POSITIVE LOGITS

netflix

0.76

ktop

0.71

yt

0.70

Dream

0.70

furt

0.69

mone

0.68

yu

0.66

mol

0.66

trak

0.66

ilipp

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.