INDEX

Explanations

architectural locations or structures

The neuron predominantly fires on numeral tokens—either digits or spelled-out numbers—identifying quantitative counts in the text.

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

 nightlife

0.71

 inmobili

0.64

房地产

0.63

 گردش

0.62

 freehold

0.62

 proprietà

0.61

 propriedade

0.61

betrieb

0.60

 вули

0.60

 தோட்டங்கள்

0.59

POSITIVE LOGITS

 hall

0.97

 pavilion

0.90

 auditorium

0.90

 Auditorium

0.89

 सभाग

0.89

 Pavilion

0.89

 Hall

0.86

 Gebäude

0.82

 Clubhouse

0.81

 classroom

0.80

Activations Density 0.429%