INDEX

Explanations

separators followed by technical identifiers

The neuron fires strongly on namespace/module separators in code—i.e. punctuation tokens like “.”, “/”, and “\” that join language or library 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

時間で

-1.34

を

-1.29

 with

-1.22

日を

-1.13

いを

-1.12

 lakini

-1.12

titutional

-1.11

曲は

-1.09

などを

-1.07

 only

-1.05

POSITIVE LOGITS

久し

1.39

 encender

1.29

 craindre

1.26

ようになって

1.25

 peinado

1.24

However

1.22

ところに

1.20

 afectar

1.20

向けの

1.19

 secado

1.18

Activations Density 0.062%