INDEX

Explanations

shared instance

The neuron strongly activates on standalone underscore tokens (often part of “g_” or other snake_case identifier prefixes) and thus is essentially detecting underscore separators in code identifiers.

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

 ausgewä

-1.76

縢

-1.69

鍆

-1.66

 höch

-1.64

 різних

-1.56

訁

-1.53

 入っ

-1.52

 Ін

-1.50

笵

-1.48

 Він

-1.41

POSITIVE LOGITS

1.61

（

1.56

</strong>

1.52

who

1.45

Они

1.40

 redondo

1.40

勠

1.40

 idée

1.39

 welches

1.38

Activations Density 0.007%