INDEX

Explanations

This neuron fires on text that belongs to distinct numbered or bulleted list items or section headings, i.e. the tokens that kick off individual list entries or outline sections.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

1.69

1.38

ä

1.30

 Solicitor

1.23

 Address

1.22

 Mustache

1.20

1.18

》

1.17

 Timeline

1.14

 Spine

1.13

POSITIVE LOGITS

ings

1.65

ungs

1.47

م

1.45

liness

1.35

و

1.31

payer

1.27

ן

1.27

ين

1.26

ية

1.26

ップ

1.24

Activations Density 28.120%