INDEX

Explanations

conclusions and technical contexts

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

-0.10

 Ã¤ven

-0.09

antz

-0.08

ç¾½

-0.08

insula

-0.08

zim

-0.08

 Colt

-0.08

ä¸Ķ

-0.08

ìĹĲëıĦ

-0.08

alama

-0.08

POSITIVE LOGITS

 therefore

0.25

 Ð¿Ð¾ÑįÑĤÐ¾Ð¼Ñĥ

0.16

 wiÄĻc

0.16

 iÅŁte

0.16

 donc

0.15

åĽłæŃ¤

0.15

 Therefore

0.14

Therefore

0.14

 quindi

0.13

 daher

0.13

Activations Density 0.123%