INDEX

Explanations

return value or result

The neuron fires on tokens that appear in code‐style API documentation or technical specs (e.g. function names, parameters, return types, callbacks, error/result descriptions).

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

 bashing

0.62

paintbrush

0.59

нинград

0.58

urent

0.57

phobic

0.56

 ప్రస్తుతం

0.55

 przeg

0.55

жнему

0.55

machen

0.54

ising

0.54

POSITIVE LOGITS

結果

1.85

结果

1.83

 result

1.80

 résultat

1.79

 resultado

1.78

 результата

1.77

 RESULT

1.77

 резулта

1.77

 Result

1.76

 results

1.73

Activations Density 0.052%