INDEX

Explanations

for purposes of

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

 Ð·Ð²Ð¸ÑĩÐ°Ð¹

-0.22

 Ð²Ð¸ÑıÐ²

-0.19

 Ð·Ð°Ð²Ð¸ÑģÐ¸Ð¼

-0.12

 Ð¸Ð½Ð´Ð¸Ð²Ð¸Ð´Ñĥ

-0.12

 Ð¿Ð¾Ð±Ð°

-0.11

 Ð²Ð¸ÑıÐ²Ð¸

-0.11

 Ð²Ð¸ÑģÑĤÑĥÐ¿

-0.10

 Ð´ÑĢÑĥÐ³Ð¸Ð¼

-0.10

enth

-0.10

POSITIVE LOGITS

 sake

0.17

 purposes

0.13

 Ð±ÑĥÐ¼Ð°

0.13

 ÑĤÐ¾Ð³Ð¾

0.13

of

0.11

Î¬Î»ÏħÏĪÎ·ÏĤ

0.11

ÑīÐµÐ³Ð¾

0.11

ÑīÐ¸Ð¹

0.11

ÑĤÐ¾Ð³Ð¾

0.11

 Mood

0.11

Activations Density 0.046%