INDEX

Explanations

saying that

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

ÐĲÑĢÑħÑĸÐ²

-0.10

ipel

-0.10

à¸Ĵ

-0.10

ofilm

-0.09

Â¿ÃĤ

-0.09

¨ë¶Ģ

-0.09

krv

-0.08

 ?><?

-0.08

#ab

-0.08

groupBox

-0.08

POSITIVE LOGITS

 also

0.21

also

0.17

 ÑĤÐ°ÐºÐ¶Ðµ

0.17

 ÑĤÐ°ÐºÐ¾Ð¶

0.14

 ëĺĲíķľ

0.13

 ÙĩÙħÚĨÙĨÛĮÙĨ

0.13

 ayrÄ±ca

0.12

 Ã©galement

0.12

 takÃ©

0.12

Also

0.12

Activations Density 0.167%