INDEX

Explanations

submission and list items

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

Negative Logits

.dtd

-0.15

 Democrats

-0.14

.decorators

-0.13

.Ð´

-0.12

thed

-0.12

.nextDouble

-0.12

slideDown

-0.12

 Denmark

-0.12

 dÃ¡v

-0.12

 dildo

-0.12

POSITIVE LOGITS

 nond

0.29

 Republicans

0.15

Republicans

0.13

dat

0.11

Republican

0.11

RNA

0.11

du

0.11

 indirect

0.11

ael

0.10

 Republican

0.10

Activations Density 0.226%

No Known Activations