INDEX

Explanations

words related to watching something, such as "watch" and "video"

oai_token-act-pair · gpt-3.5-turbo

instances of the word "watch" in relation to viewing activities

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

rez

-0.75

ļéĨĴ

-0.63

ha

-0.62

aclysm

-0.59

Merit

-0.59

INF

-0.59

olution

-0.59

 Seeds

-0.58

anes

-0.58

iguous

-0.58

POSITIVE LOGITS

 watch

3.79

watch

2.69

 watches

2.64

 Watch

2.42

 WATCH

2.09

Watch

1.99

 watched

1.96

 watching

1.83

 listen

1.55

watching

1.52

Activations Density 0.023%

words related to watching something, such as "watch" and "video"

instances of the word "watch" in relation to viewing activities

No Comments

No Known Activations