INDEX

Explanations

areas for improvement

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

strap

-0.10

chal

-0.09

uml

-0.09

ingly

-0.09

::::|

-0.09

ation

-0.08

baar

-0.08

birthdate

-0.08

oz

-0.08

urga

-0.08

POSITIVE LOGITS

 improvement

0.21

 Improvement

0.19

 areas

0.19

 where

0.16

 Areas

0.15

 improvements

0.15

 improved

0.14

 improve

0.14

areas

0.13

Areas

0.13

Activations Density 0.017%