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

iasis

-0.85

agher

-0.76

phia

-0.73

 Hitchcock

-0.70

sung

-0.68

acle

-0.65

entric

-0.64

 recess

-0.64

 Jung

-0.63

instein

-0.62

POSITIVE LOGITS

 controls

1.81

 Controls

1.13

control

0.75

Control

0.75

rules

0.69

?????-

0.68

 control

0.68

æł

0.68

ãĥīãĥ©ãĤ´ãĥ³

0.67

 buttons

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.