INDEX

Explanations

rest of

The neuron fires on the first content words of new document sections or headings—e.g. the title-style tokens that kick off a legal clause or paragraph.

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

য়েছে

0.75

 многое

0.70

場合の

0.69

人々

0.68

য়েছে

0.64

 utilizzare

0.63

 জন্যই

0.63

 ceva

0.62

 llevando

0.61

 Niveau

0.61

POSITIVE LOGITS

of

1.95

of

1.34

 ofthe

1.34

ของการ

1.27

 ऑफ

1.23

 ഓഫ്

1.21

ofthe

1.21

ofx

1.19

OfString

1.13

 ఆఫ్

1.13

Activations Density 0.000%