INDEX

Explanations

heart disease and failure

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

anna

-0.11

rous

-0.10

 body

-0.10

modo

-0.09

MainFrame

-0.09

usercontent

-0.09

nder

-0.09

itarian

-0.08

hower

-0.08

POSITIVE LOGITS

ache

0.20

burn

0.20

be

0.18

strings

0.18

broken

0.17

aches

0.17

 failure

0.17

 rate

0.17

 disease

0.16

 Rate

0.16

Activations Density 0.015%