INDEX

Explanations

specific factual points

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

emy

-0.12

oyer

-0.09

 Dread

-0.09

 ballo

-0.09

Tul

-0.08

/REC

-0.08

 Telefon

-0.08

 Below

-0.08

 ÅŁÃ¶yle

-0.08

 dece

-0.08

POSITIVE LOGITS

 note

0.23

 Note

0.20

 remember

0.17

note

0.17

Note

0.17

 These

0.16

 Remember

0.16

è®°

0.16

 these

0.16

è¿ĻäºĽ

0.15

Activations Density 0.082%