INDEX

Explanations

code delimiters and keywords

The neuron fires strongly on tokens that mark statement or block boundaries in code or prose—e.g. “break,” “;;,” “quit,” and long dash separators.

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

さも

-1.16

Արտաքին

-1.13

сон

-1.07

 střed

-1.05

 Alemanha

-1.05

 něj

-1.05

ecs

-1.05

汩

-1.05

oluble

-1.04

besondere

-1.03

POSITIVE LOGITS

 ermöglich

1.18

誌

1.16

 präsenti

1.12

 você

1.10

 nadie

1.07

1.06

，《

1.05

1.03

 Kräfte

1.02

Activations Density 0.000%