INDEX

Explanations

challenge rating

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

oub

-0.11

weg

-0.10

 meter

-0.09

 Ange

-0.09

apa

-0.08

AQ

-0.08

prefs

-0.08

 irresist

-0.08

junction

-0.08

Jag

-0.08

POSITIVE LOGITS

DC

0.16

AC

0.12

 Prof

0.12

 prof

0.11

 proficiency

0.11

dc

0.11

ycastle

0.10

 Constitution

0.10

 proficient

0.10

dc

0.10

Activations Density 0.036%