INDEX

Explanations

code block terminators or continuations

The neuron reliably fires on code snippets where variables are compared to null (e.g. “param == null” or “ptr != NULL”)—i.e. null‐check conditions in code.

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

жливо

0.58

දු

0.58

Không

0.55

່າ

0.55

まっている

0.55

Daw

0.55

ບໍ່

0.54

মোঃ

0.54

quali

0.54

フィス

0.53

POSITIVE LOGITS

</code>

1.13

 ...)

1.08

 ...]

0.99

 ....

0.99

 .....

0.93

 (...)

0.93

 …..

0.92

 ……..

0.91

 .......

0.91

 blah

0.89

Activations Density 0.006%