INDEX

Explanations

asking for help or explanation

The neuron detects auxiliary verbs and pronouns that mark the start of a question (e.g. “does,” “can,” “is,” “did,” “I” in question context).

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

//

-1.51

ara

-1.43

 möjlighet

-1.40



-1.39

-1.32

इटम

-1.28

деров

-1.26

 olabilir

-1.25

 शुर

-1.25

ové

-1.25

POSITIVE LOGITS

 here

1.69

(“

1.47

 besonderes

1.41

 today

1.27

Mga

1.27

渽

1.26

Aé

1.24

 تیم

1.22

 bricola

1.22

 broader

1.20

Activations Density 0.043%