INDEX

Explanations

defeat and losing

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

ushman

-0.11

APH

-0.10

 Bonus

-0.09

 bonus

-0.09

successfully

-0.09

bonus

-0.09

 erfolgreich

-0.09

Intialized

-0.09

 showdown

-0.09

POSITIVE LOGITS

 defeat

0.42

 losing

0.40

æķĹ

0.39

 loss

0.38

 lose

0.35

è¾ĵ

0.34

è´¥

0.33

loss

0.32

 lost

0.32

losing

0.32

Activations Density 0.086%