INDEX

Explanations

improvement and quality evaluation

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

::.

-0.11

 improvement

-0.10

Transactional

-0.09

 konce

-0.09

 pushViewController

-0.09

ViewInit

-0.09

ppers

-0.09

borg

-0.08

umb

-0.08

 operation

-0.08

POSITIVE LOGITS

 support

0.15

 communication

0.13

 reporting

0.12

 oversight

0.11

 documentation

0.11

 visibility

0.11

 Communication

0.10

æ²Ł

0.10

 scrutiny

0.10

support

0.10

Activations Density 0.057%