INDEX

Explanations

editorial notes within written text

oai_token-act-pair · gpt-3.5-turbo

references to editorial notes or acknowledgments in a text

oai_token-act-pair · gpt-4o-mini Triggered by @bot

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

 Band

-0.73

bands

-0.69

band

-0.68

 clearance

-0.68

 capacity

-0.67

sted

-0.67

 band

-0.66

 bands

-0.66

 duty

-0.65

 secure

-0.64

POSITIVE LOGITS

Editor

4.09

editor

2.27

 Editor

2.22

 Editors

1.72

 editor

1.71

Edit

1.68

 editors

1.50

 Editorial

1.46

Edited

1.43

 EDIT

1.37

Activations Density 0.017%

editorial notes within written text

references to editorial notes or acknowledgments in a text

No Comments

No Known Activations