INDEX

Explanations

defense, dirt, computer, design, travel, carbon, ice

The neuron fires on isolated, capitalized nouns used as section or product headings (e.g. “Award,” “Film,” “Salmon,” “Truck”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

bpm

0.57

含有

0.54

kart

0.53

 koleg

0.52

kaj

0.51

 longterm

0.51

jq

0.51

daq

0.51

cls

0.50

xj

0.50

POSITIVE LOGITS

Doesn

0.73

 Doesn

0.72

 gedaan

0.68

色列

0.66

ασία

0.66

はもちろん

0.63

iniai

0.63

ación

0.62

 utilisées

0.61

 Française

0.61

Activations Density 0.000%