INDEX

Explanations

should not restrict or inform

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

 ilma

-0.09

 Broad

-0.09

 provoz

-0.09

Wed

-0.08

èľ

-0.08

resents

-0.08

AllowAnonymous

-0.08

 requestCode

-0.08

licant

-0.08

Bea

-0.08

POSITIVE LOGITS

 ever

0.16

 contain

0.16

 containing

0.14

-ever

0.14

 contains

0.13

 chá»©a

0.13

 EVER

0.12

 Contains

0.12

ever

0.11

 longer

0.11

Activations Density 0.026%