INDEX

Explanations

phrases indicating user actions or interactions

oai_token-act-pair · gpt-3.5-turbo

phrases addressing the reader directly

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

 Coliseum

-0.69

AMS

-0.58

amount

-0.57

offs

-0.57

inis

-0.56

 incidental

-0.56

wall

-0.55

olid

-0.55

atown

-0.55

entimes

-0.55

POSITIVE LOGITS

 hereby

1.00

 yourselves

0.95

Reviewer

0.94

 welcome

0.83

Tube

0.83

 yourself

0.77

 browser

0.77

 guessed

0.70

isode

0.70

 invited

0.69

Activations Density 0.272%

phrases indicating user actions or interactions

phrases addressing the reader directly

No Comments

No Known Activations