INDEX

Explanations

action words that give instructions or recommendations to others

oai_token-act-pair · gpt-3.5-turbo

commands or directives urging action from the audience

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

gradient

-0.73

Orig

-0.69

croft

-0.68

quickShipAvailable

-0.62

paren

-0.62

aired

-0.62

til

-0.61

Ear

-0.61

WOR

-0.60

Cooldown

-0.59

POSITIVE LOGITS

 donate

1.25

 participate

1.21

 beware

1.16

 gather

1.15

 nominate

1.13

 partake

1.12

 submit

1.12

 undertake

1.08

 emulate

1.07

 join

1.07

Activations Density 0.158%

action words that give instructions or recommendations to others

commands or directives urging action from the audience

No Comments

No Known Activations