INDEX

Explanations

the improvement of quality

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

 responsiveness

-0.12

 suitability

-0.12

keh

-0.11

 attractiveness

-0.11

sembled

-0.09

fug

-0.09

Ð»Ð°Ð´Ð¸

-0.09

 viability

-0.09

/dashboard

-0.08

acles

-0.08

POSITIVE LOGITS

 overall

0.23

 quality

0.19

overall

0.19

 chances

0.19

 speed

0.19

 efficiency

0.17

 odds

0.16

 Overall

0.15

quality

0.15

 accuracy

0.14

Activations Density 0.050%