INDEX

Explanations

know well

The neuron activates on the adverb “well” (often in phrases like “know well,” “very well,” or “so well at…”) indicating it’s detecting uses of “well” as an intensifier or marker of proficiency/familiarity.

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

↵↵

-1.68

</em>

-1.49

</i>

-1.40

Đ

-1.34

↵

-1.34

-1.29

絝

-1.28

箪

-1.27

ztes

-1.26

ку

-1.24

POSITIVE LOGITS

 ilman

1.51

結果は

1.41

轻易

1.41

1.39

 simplif

1.39

 will

1.39

 sierpnia

1.37

 azonban

1.37

 میکنم

1.36

 zult

1.34

Activations Density 0.026%