INDEX

Explanations

references to publications and authors

oai_token-act-pair · gpt-3.5-turbo

references to sources or authors in articles or publications

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 mechanically

-0.78

ETF

-0.64

 teamwork

-0.64

henko

-0.64

 helpers

-0.63

 transfers

-0.63

 perfected

-0.63

 refunds

-0.62

 ceilings

-0.62

 hinges

-0.61

POSITIVE LOGITS

 Politico

1.31

 Huffington

1.30

 Wired

1.28

 Forbes

1.27

 Newsweek

1.23

 Tablet

1.23

 Slate

1.22

 Daily

1.22

 Vanity

1.20

 Examiner

1.19

Activations Density 0.215%

references to publications and authors

references to sources or authors in articles or publications

No Comments

No Known Activations