INDEX

Explanations

delete, remove, or kill

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

æĴ¤

-0.10

Shutdown

-0.09

Rew

-0.09

 rewind

-0.09

 Reduction

-0.09

amine

-0.09

 forfeiture

-0.09

istik

-0.09

eldorf

-0.09

kees

-0.08

POSITIVE LOGITS

 remove

0.14

 removed

0.13

rm

0.10

 removes

0.10

 drop

0.10

end

0.09

 removing

0.09

remove

0.09

 Remove

0.09

 stopping

0.09

Activations Density 0.124%

No Known Activations