INDEX

Explanations

index_table, const

This neuron fires on section‐heading or menu‐style labels (e.g. “Merch,” “Demos,” “Licensing,” “Inquiries”) that mark distinct parts of a structured document.

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

 Якщо

0.67



0.61

Якщо

0.60

 Từ

0.59

 desember

0.57

 appreciably

0.57

从

0.56

ў

0.56

Ketika

0.55

י

0.55

POSITIVE LOGITS

 etcétera

0.64

တို့

0.55

 وغیرہ

0.55

 वगैर

0.52

 enfim

0.52

 इत्यादि

0.52

etc

0.51

aka

0.51

 इंसान

0.51

usw

0.51

Activations Density 0.602%