INDEX

Explanations

demonstrating commitment or value

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

 Reputation

-0.10

 dignity

-0.10

imps

-0.09

 Gain

-0.09

-0.08

 CommandLine

-0.08

 opport

-0.08

 Accuracy

-0.08

 sane

-0.08

 Hayward

-0.08

POSITIVE LOGITS

 commitment

0.21

 intent

0.15

 intentions

0.14

 status

0.14

 support

0.13

 intention

0.13

 worth

0.13

Intent

0.13

commit

0.13

 appreciation

0.13

Activations Density 0.040%