INDEX

Explanations

returning to, Anderson documented, translators, reportedly found

academic research and classification

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

给予

0.46

𝓟

0.45

ድግዳ

0.44

isième

0.44

经常

0.43

 οργαν

0.43

мую

0.42

斉

0.41

更好地

0.41

 তোলা

0.41

POSITIVE LOGITS

0.54

 ambush

0.52

 Optional

0.50

aph

0.50

 amaze

0.49

 flair

0.48

 obsidian

0.48

 eisen

0.47

 concealment

0.47

 boss

0.47

Activations Density 0.000%