INDEX

Explanations

some additional or new

structure and transformation

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 wers

0.43

 Manning

0.41

各種

0.39

 δημο

0.39

 inégal

0.39

ธรรม

0.38

 ಇದರಿಂದ

0.38

 όμως

0.38

 opis

0.38

 olim

0.38

POSITIVE LOGITS

insertCell

0.49

 Tanık

0.43

ʔ

0.43

 яичко

0.42

[]:

0.42

 teaser

0.42

 Zahlung

0.41

 excitedly

0.41

_{

0.40

 lysate

0.40

Activations Density 0.003%