INDEX

Explanations

asking for information or help

The neuron detects question-forming words and phrases (like “if,” “how,” “what,” “know,” “where,” etc.) that signal a user is asking a question.

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

 semblent

-1.30

みたい

-1.24

 radioButton

-1.23

 бути

-1.20

 golfe

-1.19

 interro

-1.18

 devraient

-1.18

 relevante

-1.17

 manifeste

-1.17

幛

-1.16

POSITIVE LOGITS

or

1.86

 there

1.52

any

1.41

kaos

1.40

 защото

1.35

↵↵

1.33

 because

1.27

كان

1.22

utusan

1.22

 だっ

1.19

Activations Density 0.059%