INDEX

Explanations

software licenses or copyright

The neuron strongly activates on numeric literal tokens (sequences of digits) in the code.

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

Student

-0.80

student

-0.79

getSource

-0.77

iliger

-0.75

 Student

-0.75

 take

-0.75

 jatuh

-0.75

takes

-0.74

Takes

-0.74

一段

-0.70

POSITIVE LOGITS

 правила

0.66

 omnipotent

0.65

 depres

0.65

 stoves

0.64

succeeded

0.64

ヲ

0.63

ランプ

0.63

 Appliances

0.63

ナル

0.62

 enunci

0.62

Activations Density 0.091%