INDEX

Explanations

really, all, surely, might, most, exactly, should

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ajas

-0.07

anan

-0.07

illy

-0.07

 foul

-0.07

 conspir

-0.07

 caus

-0.07

laus

-0.06

 demonstr

-0.06

 Census

-0.06

POSITIVE LOGITS

 chiefly

0.12

DTV

0.11

 Perc

0.11

OID

0.11

 posto

0.10

ounce

0.10

 chaired

0.10

 unknow

0.10

 Serena

0.10

Wer

0.10

Activations Density 0.009%