INDEX

Explanations

closing remarks and greetings

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

allo

-0.10

 Hearts

-0.09

worth

-0.09

ur

-0.09

åĸľ

-0.09

echn

-0.09

Fog

-0.09

 kicking

-0.08

 plantation

-0.08

 hearts

-0.08

POSITIVE LOGITS

 regards

0.19

incerely

0.17

Best

0.16

 Best

0.16

 Cord

0.15

 regard

0.14

 cord

0.14

 kind

0.13

 best

0.12

kin

0.12

Activations Density 0.007%

No Known Activations