INDEX

Explanations

initials or abbreviations

This neuron fires on uppercase alphanumeric tokens—acronyms, abbreviations, model or class names, and other all-caps identifiers.

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

ها

-1.86

"";

-1.56

 возможность

-1.55

 ample

-1.52

 jesteśmy

-1.50

ла

-1.45

〝

-1.41

甡

-1.41

 osobow

-1.39

顰

-1.38

POSITIVE LOGITS

1.95

ﮨ

1.89

1.77

ﮦ

1.73

擽

1.66

 madness

1.64

了他的

1.63

tividade

1.61

鬃

1.61

Activations Density 0.106%