INDEX

Explanations

hexadecimal identifiers and code snippets

The neuron fires strongly on long alphanumeric identifier tokens—especially hexadecimal or numeric ID 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

and

-1.55

on

-1.41

lampe

-1.30

or

-1.30

%"),

-1.22

dekat

-1.22

in

-1.16

prnewswire

-1.13

FFEE

-1.13

 what

-1.12

POSITIVE LOGITS

が

1.22

 Ayur

1.20

 стихотво

1.17

.¹

1.15

 Irrit

1.14

 exasper

1.14

 erec

1.14

 Magen

1.14

 robus

1.13

Activations Density 0.005%