INDEX

Explanations

in embracing, she sees, fun data, he hit

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

Negative Logits

 phosph

-0.09

 Cool

-0.09

Sho

-0.09

 rough

-0.09

ETA

-0.08

 relaxation

-0.08

 relax

-0.08

 seriously

-0.08

 Soda

-0.08

POSITIVE LOGITS

 dumpster

0.14

èĳ

0.11

 cluster

0.11

 insign

0.10

avit

0.10

ke

0.09

 dysfunction

0.09

æ¨Ĥ

0.09

oron

0.09

ï¾Ĳ

0.09

Activations Density 0.078%

No Known Activations