INDEX

Explanations

add_subdirectory, شو, سب

The neuron activates on tokens written in the Arabic‐derived script (e.g. Persian/Urdu words).

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

 obliter

0.80

COMANDA

0.78

Iceland

0.78

 exacting

0.75

KeyValue

0.74

Douglas

0.73

솎

0.73

Dou

0.73

경상

0.72

Afghanistan

0.72

POSITIVE LOGITS

 های

1.28

 در

1.23

‌های

1.22

 برای

1.14

 از

1.13

 ها

1.12

 بی

1.05

 با

1.02

 این

1.02

یاز

1.02

Activations Density 0.006%