INDEX

Explanations

phrases indicating subscription or delivery of content to a recipient's inbox

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "your" in a context related to newsletters or subscriptions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

acca

-0.79

}}}

-0.71

 downs

-0.70

cific

-0.66

alys

-0.64

lda

-0.63

tymology

-0.62

stals

-0.62

cies

-0.60

uin

-0.60

POSITIVE LOGITS

 inbox

1.40

 mailbox

1.12

 clipboard

0.97

 liking

0.84

 favorite

0.81

RSS

0.79

 subscribed

0.78

 favorites

0.78

 desktop

0.76

 browser

0.75

Activations Density 0.049%

phrases indicating subscription or delivery of content to a recipient's inbox

occurrences of the word "your" in a context related to newsletters or subscriptions

No Comments

No Known Activations