INDEX

Explanations

phrases describing tendencies or inclinations

oai_token-act-pair · gpt-3.5-turbo

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

zbek

-0.68

yz

-0.63

gur

-0.62

 Polo

-0.62

loo

-0.55

pelling

-0.54

terday

-0.53

ZA

-0.52

 Slate

-0.52

fts

-0.52

POSITIVE LOGITS

rils

1.33

entious

1.24

 toward

1.14

 towards

1.07

to

1.01

ril

0.93

entimes

0.90

erest

0.84

ered

0.79

erers

0.75

Activations Density 0.035%

phrases describing tendencies or inclinations

No Comments

No Known Activations