INDEX

Explanations

scores, dates, and numbers

This neuron activates on tokens that contain multi‐digit numbers (i.e. numeric strings).

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

 fortfarande

-0.92

 doesn

-0.89

new

-0.88

 weathered

-0.86

 هنوز

-0.86

 sonunda

-0.84

 wasn

-0.84

 like

-0.84

 способом

-0.84

 głównie

-0.83

POSITIVE LOGITS

🤯

0.97

 Cien

0.93

 bewe

0.93

 versie

0.93

 SECRETARY

0.87

NOWLEDGE

0.86

 CORPOR

0.86

RouterModule

0.86

manned

0.84

 WATERS

0.84

Activations Density 0.003%