INDEX

Explanations

triple-quoted strings

The neuron fires on terms and verbs used in programming‐style documentation or changelogs—i.e. words that introduce or describe code features and grammar (like “Add,” “Support,” “raw,” “regex,” “splitting,” etc.).

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

 குறிப்பிட்டு

0.56

 मंड

0.55

বৎ

0.54

獲得

0.53

 permissive

0.52

 preventable

0.52

.-(

0.51

 গৌরব

0.50

 मिळ

0.50

 vener

0.50

POSITIVE LOGITS

0.89

0.80

"""

0.73

``

0.71

)`

0.68

\`

0.66

>`

0.65

`;

0.64

``

0.64

`,

0.62

Activations Density 0.003%