INDEX

Explanations

acknowledging limitations and complexities via conjunctions

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

 setbacks

-0.09

Ã

-0.09

rok

-0.09

 testimony

-0.08

iran

-0.08

 tandem

-0.08

 cÃ¡c

-0.08

 Ð½Ð°ÑĪÐ¸Ñħ

-0.08

\OptionsResolver

-0.08

POSITIVE LOGITS

 conven

0.11

'gc

0.11

/or

0.10

Â´

0.10

 invol

0.10

ruptions

0.10

 clar

0.09

 secre

0.09

ones

0.09

 simp

0.09

Activations Density 0.132%