INDEX

Explanations

the word "on" in various contexts

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate staying in a certain state or position

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

YR

-0.67

estic

-0.63

»Ĵ

-0.63

UGE

-0.63

UTF

-0.63

erity

-0.63

ĪĴ

-0.62

DragonMagazine

-0.61

ians

-0.59

catentry

-0.58

POSITIVE LOGITS

 indefinitely

0.78

lishes

0.75

 haunt

0.72

quo

0.69

kos

0.68

 confines

0.67

 haun

0.67

 afloat

0.66

 sidelines

0.65

tracks

0.64

Activations Density 0.103%

the word "on" in various contexts

phrases that indicate staying in a certain state or position

No Comments

No Known Activations