INDEX

Explanations

ways and manner

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

 juxtap

-0.10

itm

-0.10

aben

-0.09

pta

-0.09

 taste

-0.09

COPE

-0.09

/by

-0.09

.scalablytyped

-0.09

edl

-0.08

ableViewController

-0.08

POSITIVE LOGITS

 ways

0.47

way

0.34

 manner

0.27

ways

0.26

 Ways

0.25

 faÃ§on

0.24

 cÃ¡ch

0.24

 extent

0.23

 maniÃ¨re

0.23

æĸ¹å¼ı

0.23

Activations Density 0.076%