INDEX

Explanations

code structure defining properties

This neuron fires on action‐oriented verbs (especially imperatives and gerunds) that signal user intent or calls to act (e.g. “forget,” “searching,” “looking”).

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

ченной

0.90

ченные

0.86

чкой

0.84

ciones

0.84

 completos

0.83

aunque

0.82

ларды

0.80

column

0.80

бы

0.80

ciem

0.79

POSITIVE LOGITS

 adatto

0.83

ী

0.75

 altro

0.73

 dahulu

0.73

0.72

mia

0.72

 familiare

0.72

{~

0.71

 interesting

0.71

 suç

0.71

Activations Density 0.001%