INDEX

Explanations

continually seek opportunities to improve

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

ï½Ľ

-0.10

Await

-0.09

 Persistent

-0.09

 persistent

-0.09

 sticky

-0.09

 adept

-0.09

 control

-0.09

ottle

-0.09

 heritage

-0.08

ontrol

-0.08

POSITIVE LOGITS

 staying

0.21

 stay

0.20

 Stay

0.18

 stays

0.18

Stay

0.17

 keeping

0.16

 update

0.16

 stayed

0.16

 keep

0.15

 keeps

0.15

Activations Density 0.066%