INDEX

Explanations

either or some kind of investigation

The neuron fires on long abstract “-ion/-ment/-ance” style nouns (formal, legal, or technical terms).

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

 всички

0.99

 EVERY

0.95

 Semua

0.94

 semua

0.94

 toate

0.90

 každý

0.90

 EVERYTHING

0.88

 všetky

0.88

 tutte

0.84

 všechny

0.84

POSITIVE LOGITS

 either

1.27

either

1.23

某种

1.07

หรือ

1.06

或其他

1.05

 или

1.03

 alguna

1.03

Either

1.02

 либо

1.00

 hoặc

0.98

Activations Density 0.019%