INDEX

Explanations

non-english strings

The neuron strongly activates on markup elements and code‐like tokens—especially XML/HTML tags, angle brackets, attribute or element names, and other structured metadata.

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

 Diarsipkan

-1.36

 Dokter

-1.25

 october

-1.23

:{},

-1.21

 Arxivat

-1.20

Якщо

-1.20

Ótimo

-1.19

OREM

-1.17

 werkt

-1.16

 groet

-1.16

POSITIVE LOGITS

it

1.57

ละเอ

1.38

のかも

1.36

 это

1.27

 теперь

1.21

他已经

1.14

nha

1.13

已被

1.13

dział

1.12

 mittelalter

1.11

Activations Density 0.007%