INDEX

Explanations

ability to act

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

 Cant

-0.15

 Canonical

-0.15

canonical

-0.15

 canonical

-0.14

 Cann

-0.13

Canonical

-0.13

 Canterbury

-0.13

cant

-0.13

 canned

-0.13

 cann

-0.12

POSITIVE LOGITS

can

0.47

åı¯ä»¥

0.44

 dapat

0.36

 kÃ¶nnen

0.30

ï¼Įåı¯ä»¥

0.30

 kunnen

0.29

à¸ªà¸²à¸¡à¸²à¸£à¸ĸ

0.29

 could

0.29

 à¤¸à¤ķà¤¤

0.29

 Ð¼Ð¾Ð¶Ð½Ð¾

0.28

Activations Density 0.133%