INDEX

Explanations

phrases related to things being published or distributed

oai_token-act-pair · gpt-3.5-turbo

the word "out" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

cious

-0.72

antry

-0.72

cius

-0.71

iosity

-0.63

 examiner

-0.60

inski

-0.60

ettlement

-0.59

antine

-0.59

 deprivation

-0.56

 etched

-0.56

POSITIVE LOGITS

lier

0.93

fitted

0.91

stretched

0.88

casts

0.85

posts

0.83

wards

0.83

doors

0.83

lander

0.80

)=(

0.80

smart

0.80

Activations Density 0.110%

phrases related to things being published or distributed

the word "out" in various contexts

No Comments

No Known Activations