INDEX

Explanations

structure and format

discussing 'however'/'but' contrast

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

κ

0.49

 टूटा

0.49

ಎ

0.48

ک

0.47

नहीं

0.46

ου

0.46

ക്കെ

0.46

 ഉയർന്ന

0.46

 Initialize

0.44

좇

0.44

POSITIVE LOGITS

 brewing

0.48

ษ

0.48

让人

0.45

 handy

0.45

 Pepper

0.43

 Ávila

0.43

 pepper

0.42

茶

0.42

 darkest

0.41

 shadowy

0.41

Activations Density 0.001%