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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

intendent

-0.97

unta

-0.76

ombies

-0.76

uga

-0.74

umbn

-0.73

Reviewer

-0.72

che

-0.72

uctor

-0.71

opsy

-0.71

hower

-0.70

POSITIVE LOGITS

 Euros

0.68

 Vide

0.67

 Eden

0.65

 Xuan

0.63

 mmol

0.62

 antioxid

0.61

 Amid

0.60

{\

0.59

pse

0.59

 privacy

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.