INDEX

Explanations

evaluate the damage, identify diseases

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

ver

-0.10

 verr

-0.09

uky

-0.09

Sir

-0.09

Circular

-0.08

 proven

-0.08

.quiz

-0.08

alin

-0.08

Sir

-0.08

 termin

-0.08

POSITIVE LOGITS

 presence

0.19

 Presence

0.15

 suspected

0.15

Presence

0.14

presence

0.14

 suspicions

0.14

 extent

0.12

 rule

0.12

 status

0.12

 signs

0.12

Activations Density 0.025%