INDEX

Explanations

closing bracket or null value

The neuron attends to code-specific tokens—identifiers, symbols, and constants (e.g. syscall labels like “SB,” variable names, hex literals, and other non-natural-language tokens).

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

for

-1.67

of

-1.43

all

-1.34

on

-1.24

 most

-1.19

to

-1.18

 first

-1.14

in

-1.13

 started

-1.13

by

-1.12

POSITIVE LOGITS

 když

1.20

ebles

1.13

 '.',

1.11

hicule

1.09

mathrm

1.09

܇

1.09

 wren

1.09

 lits

1.08

receta

1.06

 cápsulas

1.05

Activations Density 0.001%