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

bot

-0.67

 polls

-0.65

 transsexual

-0.61

 Clair

-0.60

Remove

-0.60

 polling

-0.59

hillary

-0.59

istor

-0.59

 organs

-0.58

 ILCS

-0.58

POSITIVE LOGITS

taboola

0.92

etheless

0.90

 also

0.87

also

0.78

ategory

0.78

isode

0.77

ngth

0.76

 nodd

0.75

 livest

0.73

ailability

0.72

Activations Density 0.000%

No Known Activations

This feature has no known activations.