INDEX

Explanations

The neuron activates on list-style introduction words—nouns like “thing(s),” “tip(s),” “question(s),” “factor(s),” etc.—that signal enumerated or advice headings.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

之时

0.61

 바탕

0.57

뭡

0.55

和我

0.54

的情况下

0.51

}((

0.50

டுவத

0.48

 repente

0.48

だから

0.47

之际

0.46

POSITIVE LOGITS

you

1.45

ที่จะ

1.43

 that

1.42

that

1.41

we

1.40

ที่คุณ

1.28

 мы

1.26

 youll

1.25

 που

1.24

你可以

1.24

Activations Density 0.323%

No Known Activations