INDEX

Explanations

closing brackets, quotes, or lists

The neuron strongly activates on non-standard or “markup” tokens—things like citation markers (e.g. [* RXTE *], {}), embedded symbols (®, —), and other bracketed or formatted metadata 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

Empleado

-1.27

 minimalista

-1.24

 lesquelles

-1.20

 länge

-1.17

 作り方

-1.14

 silikon

-1.13

Étymologie

-1.13

Leírás

-1.13

Cevap

-1.13

Uwagi

-1.12

POSITIVE LOGITS

who

1.34

to

1.30

aphone

1.15

apropri

1.09

is

1.08

estries

1.04

уме

1.03

：……

1.02

/{

1.02

na

1.01

Activations Density 0.022%