INDEX

Explanations

may feel, colleagues are, may have

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

.scalablytyped

-0.11

bai

-0.10

çĸĳ

-0.09

-cols

-0.09

 -âĢĲ

-0.09

 almost

-0.09

 seem

-0.09

 potentially

-0.09

 seems

-0.09

 somewhat

-0.09

POSITIVE LOGITS

 simply

0.42

 Simply

0.29

 Ð¿ÑĢÐ¾ÑģÑĤÐ¾

0.29

 simplement

0.29

Simply

0.27

 einfach

0.25

 prostÄĽ

0.24

sim

0.18

 perhaps

0.16

æł¹æľ¬

0.16

Activations Density 0.096%