INDEX

Explanations

notes or annotations within text

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Note" or variations of it

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

gut

-0.73

 quit

-0.65

 hust

-0.64

 tactics

-0.63

 volunteers

-0.63

pal

-0.62

 fight

-0.61

 open

-0.61

ulic

-0.61

adian

-0.60

POSITIVE LOGITS

Note

3.48

 Note

2.33

note

2.29

NOTE

2.28

Notes

1.94

Notice

1.86

 NOTE

1.83

 note

1.64

Important

1.53

Warning

1.47

Activations Density 0.018%

notes or annotations within text

instances of the word "Note" or variations of it

No Comments

No Known Activations