INDEX

Explanations

improving human health

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

ï¾ı

-0.09

emb

-0.09

 Pleasant

-0.09

 Hunger

-0.09

 barg

-0.09

CERT

-0.08

 nurse

-0.08

pto

-0.08

tero

-0.08

 Crown

-0.08

POSITIVE LOGITS

 lives

0.17

 saving

0.17

 Lives

0.15

 saves

0.15

 patient

0.15

 treatments

0.14

 save

0.13

 salvar

0.13

 treat

0.13

saving

0.13

Activations Density 0.069%