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

oat

-0.75

itored

-0.72

 blacklist

-0.69

omin

-0.67

speak

-0.66

aqu

-0.66

unal

-0.65

hyp

-0.64

via

-0.64

olkien

-0.64

POSITIVE LOGITS

 Mechdragon

0.72

 SPORTS

0.66

 Townsend

0.65

Dug

0.63

esville

0.63

leground

0.63

--+

0.62

 Democr

0.62

 Flavoring

0.61

 Colo

0.60

Activations Density 0.000%

No Known Activations

This feature has no known activations.