INDEX

Explanations

implications and complementary perspectives

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

 Dudley

-0.09

.DataContext

-0.09

³»

-0.09

angan

-0.09

 Wonderland

-0.08

 HÃ¼s

-0.08

 conven

-0.08

 BoxFit

-0.08

 defining

-0.08

 advent

-0.08

POSITIVE LOGITS

 applicable

0.17

 help

0.15

 useful

0.15

 applied

0.14

 helps

0.14

 implications

0.13

help

0.13

-help

0.13

 insights

0.13

 applies

0.12

Activations Density 0.084%