INDEX

Explanations

code-related terms and legal citations

The neuron responds to uncommon, content-bearing words—especially technical terms, proper nouns, and other low-frequency 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

 Knight

-0.90

Knight

-0.82

ದು

-0.73

 bottom

-0.71

MMS

-0.71

KEM

-0.68

Sum

-0.68

罪

-0.68

 bower

-0.68

PHE

-0.67

POSITIVE LOGITS

 befre

0.70

inha

0.69

 اک

0.68

 中國

0.66

 PhpStorm

0.65

 reiz

0.64

 marcar

0.64

 Schwan

0.63

outline

0.63

 permette

0.63

Activations Density 0.047%