INDEX

Explanations

depends on factors

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

ê°¤

-0.10

uest

-0.09

ted

-0.09

 ÑĸÐ½ÑĪÐ¾Ð³Ð¾

-0.09

acha

-0.09

LTR

-0.08

emmel

-0.08

 chac

-0.08

ties

-0.08

ubyte

-0.08

POSITIVE LOGITS

 several

0.20

Several

0.15

 Several

0.14

 beberapa

0.14

åĩłä¸ª

0.13

 somewhat

0.13

 ëªĩ

0.13

veral

0.12

few

0.12

 Ð½ÐµÑģÐºÐ¾Ð»ÑĮÐºÐ¾

0.12

Activations Density 0.095%