INDEX

Explanations

phrases indicating a question or doubt about something

oai_token-act-pair · gpt-3.5-turbo

phrases that express uncertainty or inquiry regarding decisions

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

Torrent

-0.59

ago

-0.57

osp

-0.56

que

-0.55

ame

-0.54

ront

-0.54

onew

-0.53

stro

-0.53

ppo

-0.53

+)

-0.52

POSITIVE LOGITS

 whether

3.18

whether

2.73

 Whether

1.95

Whether

1.89

how

1.34

 regardless

1.31

 irrespective

1.27

why

1.26

if

1.10

 whence

1.01

Activations Density 0.025%

phrases indicating a question or doubt about something

phrases that express uncertainty or inquiry regarding decisions

No Comments

No Known Activations