INDEX

Explanations

code generation delimiters or prefixes

The neuron fires on numeric tokens—especially version or ID numbers embedded in code comments and file‐path strings.

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

 this

-1.88

 would

-1.86

all

-1.70

 could

-1.59

 both

-1.57

 some

-1.49

 over

-1.48

had

-1.30

 with

-1.30

licante

-1.25

POSITIVE LOGITS

 hornear

1.31

 různých

1.30

 empêcher

1.27

 douze

1.26

有时

1.22

なるべく

1.21

 verschiedener

1.20

 diverses

1.20

 manières

1.18

беріга

1.17

Activations Density 0.006%