INDEX

Explanations

This neuron fires on content words naming abstract or thematic concepts—especially nouns like battle, trust, victory, integrity, perseverance, curiosity, patterns, reasons, etc.

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

 είναι

0.69

 memiliki

0.56

 adalah

0.53

 является

0.52

 wikipagina

0.49

 ৩০

0.49

 показатель

0.49

 ১০

0.49

 tener

0.49

 مي‌

0.48

POSITIVE LOGITS

and

0.84

and

0.63

และ

0.57

性和

0.53

to

0.52

for

0.52

 અને

0.52

和

0.51

if

0.48

andre

0.48

Activations Density 0.004%