INDEX

Explanations

the central

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

................................................................

-0.14

 ................................................................

-0.13

ÂŃing

-0.11

èĤ¡ä»½æľīéĻĲåħ¬åı¸

-0.11

ç§ĳæĬĢæľīéĻĲåħ¬åı¸

-0.11

-0.10

................................

-0.10

 straightforward

-0.09

ÂłÂł

-0.09

 ................................

-0.09

POSITIVE LOGITS

&\n

0.17

 misunder

0.14

 Ð¿ÑĢÐ°ÐºÑĤÐ¸

0.12

 prostitut

0.10

 ÑĩÐµÐ»Ð¾Ð²

0.10

èĤ¡ä»½

0.10

ãĢģ\n

0.09

ï¼Į\n

0.09

,\n

0.09

 *,\n

0.09

Activations Density 0.223%