INDEX

Explanations

pairs joined by 'and' or 'et'

New Auto-Interp

Top Features by Cosine Similarity

Configuration

Prompts (Dashboard)

10,000 prompts, 128 tokens each

Dataset (Dashboard)

lmsys/lmsys-chat-1m

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Äįi

-0.09

(and

-0.08

 à¹Ĩ

-0.08

åıĬ

-0.08

 FIXME

-0.08

ÃŃr

-0.08

etc

-0.07

etc

-0.07

ych

-0.07

çĶļ

-0.07

POSITIVE LOGITS

 åĴĮ

0.18

 ä¸İ

0.16

ãĤĵãģ¨

0.15

 ãģ¨

0.14

and

0.14

íķĻê³¼

0.13

ãĤĬãģ¨

0.13

ìĻĢ

0.13

 vÃł

0.13

åĴĮ

0.13

Activations Density 0.212%