INDEX

Explanations

by sending, same subject, been degraded, I make, the rules

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

 ÐŀÐ´Ð½Ð°

-0.17

ÐĶÐĲ

-0.12

ovic

-0.12

ÑģÐ¸Ð»ÑĮ

-0.12

Âłje

-0.12

 ÐĹÐ°ÑĤÐµÐ¼

-0.11

 ÐŁÑĢÐ°

-0.11

Ð¿ÑĢÐ¸Ð¼ÐµÑĢ

-0.11

Ð·ÑĮ

-0.11

 ÐĺÑħ

-0.11

POSITIVE LOGITS

âĦĸâĦĸ

0.24

Ñģ

0.22

 Ñģ

0.21

 Ð¡

0.20

Ð¡

0.19

Ð°

0.18

 âĦĸ

0.18

Ñĥ

0.17

âĦĸ

0.16

Ð¾

0.16

Activations Density 0.300%