INDEX

Explanations

questions starting with "How can"

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hig

-0.82

renheit

-0.72

yssey

-0.68

xon

-0.67

 teasp

-0.63

iHUD

-0.63

assador

-0.63

psc

-0.62

chnology

-0.62

cedented

-0.62

POSITIVE LOGITS

 anyone

1.05

you

1.00

we

0.97

 anybody

0.96

 someone

0.95

 somebody

0.86

 they

0.86

't

0.82

YOU

0.77

ya

0.72

Activations Density 0.064%

questions starting with "How can"

No Comments

No Known Activations