INDEX

Explanations

exploit, abuse, or endanger

The neuron fires on individual list‐item words—e.g. side‐effect symptoms or legal entities—that appear in comma‐separated warning or disclaimer lists.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ভিতে

0.45

ції

0.40

 лінії

0.40

}$.

0.39

вець

0.39

➧

0.38

 sehingga

0.38

এরপর

0.38

ủa

0.38

那时候

0.37

POSITIVE LOGITS

などを

0.65

and

0.63

etc

0.62

等人

0.61

 மற்றும்

0.59

 และ

0.59

等的

0.59

 మరియు

0.58

などが

0.58

 និង

0.57

Activations Density 0.058%