INDEX

Explanations

with challenges

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

 mistr

-0.10

 bitterness

-0.09

 desperation

-0.09

pul

-0.09

otic

-0.09

Ã©f

-0.08

 hardship

-0.08

itar

-0.08

 stup

-0.08

 tort

-0.08

POSITIVE LOGITS

 homes

0.21

 Homes

0.18

homes

0.16

 pressure

0.15

 perfection

0.15

 impost

0.15

 burn

0.15

 Pressure

0.14

jet

0.14

pressure

0.14

Activations Density 0.128%