INDEX

Explanations

refers to that

This neuron fires on imperative “command” words (e.g. show, do, set, focus) that appear in technical how-to or code-instruction contexts.

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

 який

-1.50

Ȝ

-1.37

Какие

-1.34

 яка

-1.33

 насеље

-1.33

Geschiedenis

-1.32

oczes

-1.30

してます

-1.29

 皇族

-1.27

ಇ

-1.27

POSITIVE LOGITS

 that

11.81

That

5.41

 THAT

5.16

 That

4.97

that

4.28

 того

4.22

その

4.22

 ذلك

4.16

นั้น

4.06

那

4.00

Activations Density 0.872%