INDEX

Explanations

please write or think

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

 pardon

-0.11

å§ĵ

-0.10

tit

-0.10

 geschichten

-0.09

 apology

-0.09

asher

-0.09

ardy

-0.09

 excuse

-0.09

eyen

-0.09

åĢĴ

-0.09

POSITIVE LOGITS

 please

0.31

please

0.28

Please

0.27

 Please

0.26

 PLEASE

0.23

PLEASE

0.19

 bitte

0.18

ï¼Įè¯·

0.18

.Please

0.17

è¯·

0.16

Activations Density 0.033%