INDEX

Explanations

names of individuals

oai_token-act-pair · gpt-3.5-turbo

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 plurality

-0.63

 simultane

-0.62

 footing

-0.60

 bandits

-0.60

 rule

-0.60

 overwhelming

-0.60

 defaults

-0.59

dogs

-0.59

 endeavour

-0.59

 turnover

-0.59

POSITIVE LOGITS

Premium

0.96

Contribut

0.91

Executive

0.90

Assistant

0.89

Legendary

0.85

join

0.80

?????-?????-

0.80

CEO

0.80

Writer

0.78

Senior

0.78

Activations Density 0.097%

names of individuals

No Comments

No Known Activations