INDEX

Explanations

metrics and their implications

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

statt

-0.10

 Terr

-0.10

 Whitney

-0.10

ember

-0.09

wis

-0.08

acci

-0.08

 Bris

-0.08

áº±

-0.08

 Noir

-0.08

POSITIVE LOGITS

 means

0.22

 generally

0.18

æĦıåĳ³

0.18

 Means

0.16

means

0.16

ãģ»ãģ©

0.13

Means

0.13

 than

0.13

à¸«à¸¡à¸²à¸¢

0.12

equ

0.12

Activations Density 0.077%