INDEX

Explanations

numerical indicators of the order of written content

oai_token-act-pair · gpt-3.5-turbo

references to specific chapters 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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

oggles

-0.76

zzle

-0.74

fitting

-0.71

pes

-0.69

berman

-0.67

yg

-0.65

zzi

-0.63

enez

-0.63

claimed

-0.63

otal

-0.62

POSITIVE LOGITS

 chapter

0.99

 chapters

0.94

ĸļ

0.90

apeake

0.84

chapter

0.83

 Chapter

0.72

 GOODMAN

0.71

 Chapters

0.71

acters

0.71

0.70

Activations Density 0.004%

numerical indicators of the order of written content

references to specific chapters in a text

No Comments

No Known Activations