INDEX

Explanations

inflate, advance, data

This neuron fires on code‐style tokens—especially variable names, object references, and method/field accesses (e.g. “binding,” “view,” “itemView,” “DataBindingUtil,” and the “.” operator) in programming snippets.

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

 only

-1.74

 используют

-1.66

 Memper

-1.56

蛩

-1.52

 from

-1.52

 нашем

-1.51

 Lainnya

-1.50

-1.49

 сказали

-1.47

 Einfluß

-1.45

POSITIVE LOGITS

1.79

‚

1.77

\_

1.73

錵

1.70

1.55

«

1.52

夘

1.52

今年の

1.52

 compels

1.52

vereignty

1.51

Activations Density 0.002%