INDEX

Explanations

accusations pro-Ukrainian, between the, how Russia

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

 Benghazi

-0.13

 Kurdistan

-0.11

 Syrians

-0.10

 Gaza

-0.10

 Raqqa

-0.10

 Iranians

-0.10

 Palest

-0.10

 Friedrich

-0.10

raquo

-0.09

Iraq

-0.09

POSITIVE LOGITS

 Washington

0.24

 President

0.21

 China

0.21

Washington

0.19

 Russia

0.18

 India

0.18

 Beijing

0.16

 Chinese

0.16

 Prime

0.15

 Moscow

0.15

Activations Density 0.125%