INDEX

Explanations

both and

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

 Bain

-0.09

Å¡ka

-0.09

 zwar

-0.09

resco

-0.09

à¹ģà¸¥à¸°à¸ģà¸²à¸£

-0.09

tant

-0.08

iaux

-0.08

uraa

-0.08

raz

-0.08

ï¼ıï¼ı

-0.08

POSITIVE LOGITS

 Ð¾Ð´Ð½Ð¾Ð²

0.30

åĲĮæĹ¶

0.28

åĲĮæĻĤ

0.28

 alike

0.26

 zÃ¡roveÅĪ

0.25

 rolled

0.21

zug

0.20

 simult

0.19

 simultaneously

0.19

 gleich

0.19

Activations Density 0.125%