INDEX

Explanations

levels, styles, categories, periods

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

å¬

-0.10

 governmental

-0.09

 pÃºblica

-0.09

(of

-0.09

enk

-0.09

 categorical

-0.09

ordon

-0.08

 Slide

-0.08

 ghetto

-0.08

 wield

-0.08

POSITIVE LOGITS

 Ø²ÙħØ§ÙĨÛĮ

0.14

de

0.12

 Ú©Ø§Ø±ÛĮ

0.11

 ØºØ°Ø§ÛĮÛĮ

0.10

 editorial

0.10

oning

0.09

cul

0.09

Ã¼gen

0.09

 pracy

0.09

atorial

0.09

Activations Density 0.172%