INDEX

Explanations

manner of, to, or while

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

æ¯«

-0.14

 exactly

-0.13

 identical

-0.12

 exact

-0.11

å®Įåħ¨

-0.11

 perfect

-0.11

Exactly

-0.10

 completely

-0.10

actly

-0.10

Exact

-0.10

POSITIVE LOGITS

 somewhat

0.23

 pretty

0.22

 SOME

0.22

 ìĸ´ëĬĲ

0.21

 reasonably

0.21

 relatively

0.21

 largely

0.20

 fairly

0.20

èĩ³å°ĳ

0.19

 generally

0.19

Activations Density 0.412%