INDEX

Explanations

error and warning messages

The neuron fires on identifiers or tokens that denote error‐ or warning‐level logging or diagnostics (e.g. “error,” “warn,” “fatal,” 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

-1.62

 still

-1.55

制服

-1.53

šil

-1.51

 íntegra

-1.51

ちなみに

-1.49

兑换

-1.46

 murió

-1.44

 díky

-1.41

 uninterrupted

-1.40

POSITIVE LOGITS

mientras

1.52

 incrí

1.49

﨑

1.47

of

1.46

 ikke

1.45

棡

1.45

 手順

1.42

 edifici

1.40

1.38

Activations Density 0.003%