INDEX

Explanations

be a great way

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

gad

-0.09

IGHT

-0.08

 Freddie

-0.08

errMsg

-0.08

 BITTE

-0.08

 sovere

-0.08

agher

-0.08

ForMember

-0.08

isos

-0.08

quisite

-0.08

POSITIVE LOGITS

 great

0.21

 excellent

0.19

way

0.17

great

0.17

 ways

0.16

 Excellent

0.13

Excellent

0.12

 excelente

0.12

 another

0.12

 GREAT

0.12

Activations Density 0.039%