INDEX

Explanations

can be used to

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

-0.12

 Gale

-0.10

 æĴ

-0.10

 optimizations

-0.09

 Optimization

-0.09

 Saga

-0.09

 Serving

-0.09

ussy

-0.08

rame

-0.08

stalk

-0.08

POSITIVE LOGITS

 capture

0.15

 handle

0.15

 robust

0.14

 handling

0.13

Rob

0.13

rob

0.13

 handles

0.13

 flex

0.12

 automatic

0.12

 prov

0.12

Activations Density 0.051%