INDEX

Explanations

Chinese culture and media

The neuron fires on tokens that are capitalized or numeric—i.e. proper names, acronyms, dates, and other uppercase or digit‐starting tokens.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

igslist

-0.89

 YouTube

-0.85

勐

-0.82

обу

-0.81

🛎

-0.77

sia

-0.77

kut

-0.75

 WhatsApp

-0.71

глед

-0.70

cstdlib

-0.69

POSITIVE LOGITS

 Mengh

0.91

десь

0.88

 refuted

0.85

考

0.84

 Searle

0.84

ürk

0.83

 orthodon

0.82

 Bail

0.82

并且

0.81

Anar

0.81

Activations Density 0.006%