INDEX

Explanations

phrases relating to titles or headings that convey important information

oai_token-act-pair · gpt-3.5-turbo

instances of the text format indicating a new section or categorization in the document

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

aea

-0.71

perties

-0.71

lets

-0.69

alk

-0.68

vae

-0.67

nels

-0.65

eca

-0.65

atron

-0.64

auga

-0.64

chuk

-0.64

POSITIVE LOGITS

 INFORMATION

1.47

LINE

1.46

HEAD

1.46

BOOK

1.46

THE

1.45

WORK

1.44

 REPORT

1.44

 PHOTO

1.43

 ABOUT

1.40

 WORK

1.39

Activations Density 0.202%

phrases relating to titles or headings that convey important information

instances of the text format indicating a new section or categorization in the document

No Comments

No Known Activations