INDEX

Explanations

instructing how to say unknown

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

inea

-0.09

Â£Â£

-0.09

voke

-0.08

ÎµÏĢ

-0.08

amar

-0.08

å®¾

-0.08

ab

-0.08

cri

-0.08

olle

-0.08

chin

-0.08

POSITIVE LOGITS

 simply

0.17

 Simply

0.15

Simply

0.14

 thÃ¬

0.13

 then

0.13

 writ

0.13

åĪĻ

0.12

 Ð¿ÑĢÐ¾ÑģÑĤÐ¾

0.12

 saying

0.12

åīĩ

0.12

Activations Density 0.062%