INDEX

Explanations

is checks (like is_array)

The neuron fires strongly on identifiers beginning with “is_” (or “is”) that name type‐checking or predicate functions.

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

を

-1.63

könig

-1.50

Zubereitung

-1.43

 سایر

-1.43

 situe

-1.42

 唯美

-1.41

また

-1.41

魆

-1.39

 dipakai

-1.39

 constate

-1.38

POSITIVE LOGITS

It

1.95

1.79

為に

1.45

During

1.45

After

1.44

1.41

ほら

1.40

1.38

Activations Density 0.002%