INDEX

Explanations

phrases related to receiving top stories or information

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

phrases related to subscription or receiving newsletters

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

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

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_post

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

iture

-0.63

 disagrees

-0.59

fan

-0.57

leaf

-0.57

dad

-0.56

rees

-0.56

ãĥĭ

-0.56

fet

-0.56

anic

-0.55

POSITIVE LOGITS

 Tablet

0.81

DAY

0.78

esson

0.65

brance

0.65

otrop

0.65

morning

0.63

ibur

0.59

antz

0.59

 Flavoring

0.58

 interesting

0.56

Activations Density 0.027%

phrases related to receiving top stories or information

phrases related to subscription or receiving newsletters

No Comments

No Known Activations