INDEX

Explanations

code snippets and keywords

The neuron strongly activates on programming identifiers—especially type‐ or class‐style names (e.g. CamelCase or code tokens like “Dependency”, “Test”, “Base”, “Event”, “Counter”, etc.).

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

),

-0.90

Greenhouse

-0.84

橢

-0.82

设施

-0.81

емного

-0.79

 rios

-0.78

 personnalisé

-0.77

 some

-0.77

Caribbean

-0.77

Bav

-0.76

POSITIVE LOGITS

between

0.78

居然

0.76

while

0.74

ഛ

0.74

/>}

0.74

zur

0.73

竟

0.73

竟然

0.72

 unveils

0.72

 fila

0.71

Activations Density 0.083%