INDEX

Explanations

superheroes and villains

The neuron fires on specific capitalized names and titles—particularly character names and film titles in entertainment contexts.

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

 پیغم

0.56

 fringes

0.54

 étn

0.54

 туристи

0.53

 peripheral

0.52

 спор

0.52

ăț

0.52

 स्थ

0.52

INR

0.52

 Danube

0.51

POSITIVE LOGITS

Superman

0.89

Ext

0.87

 Superman

0.87

Batman

0.84

 Batman

0.84

 superman

0.83

Ext

0.83

 batman

0.81

superclass

0.73

 Gotham

0.70

Activations Density 0.056%