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

 detrim

-0.73

 insurg

-0.70

 metic

-0.69

 Tend

-0.68

busters

-0.66

 galaxies

-0.64

Ips

-0.64

 domination

-0.63

 dissent

-0.63

psych

-0.63

POSITIVE LOGITS

ACTED

0.72

ragon

0.72

>[

0.72

iere

0.71

/-

0.70

OULD

0.69

legram

0.67

gat

0.67

ilant

0.66

hran

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.