INDEX

Explanations

how you can learn

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

 háºŃu

-0.10

ydk

-0.10

umer

-0.09

alue

-0.09

\Dependency

-0.08

_PC

-0.08

éĹ´

-0.08

erville

-0.08

Dismiss

-0.08

oba

-0.08

POSITIVE LOGITS

 learn

0.19

 Learn

0.15

 learning

0.15

 self

0.14

 learns

0.14

learn

0.13

Learn

0.12

learning

0.12

.learn

0.12

-learning

0.11

Activations Density 0.196%