INDEX

Explanations

the followed by a noun

This neuron activates on words that are clearly non-English—i.e. tokens belonging to foreign-language 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

ið

0.75

 eleições

0.70

 diciembre

0.70

 noviembre

0.69

 novembro

0.67

 layoffs

0.64

 enero

0.64

 España

0.63

 sonrası

0.63

ılan

0.62

POSITIVE LOGITS

まず

0.59

↣

0.58

 dexter

0.55

GetOffset

0.53

 selben

0.53

बहुत

0.52

によって

0.52

InsertInt

0.52

 BLUENRG

0.51

သူ

0.50

Activations Density 0.152%