INDEX

Explanations

words or phrases related to watching or viewing content, such as "Watch" followed by a video title or description

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Watch."

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ctrl

-0.78

pex

-0.73

 misunderstanding

-0.66

ãĤ¨ãĥ«

-0.66

interstitial

-0.65

acqu

-0.65

 bleeding

-0.64

 blunt

-0.63

 academia

-0.62

xual

-0.62

POSITIVE LOGITS

 Watching

1.28

Watch

1.26

tower

1.19

 Watch

1.15

watch

1.07

 WATCH

0.92

Monitor

0.86

dog

0.83

 watch

0.83

wat

0.83

Activations Density 0.008%

words or phrases related to watching or viewing content, such as "Watch" followed by a video title or description

instances of the word "Watch."

No Comments

No Known Activations