could be, include, potentially

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

 usually

-0.10

''''

-0.09

ned

-0.09

uela

-0.09

azen

-0.08

POSITIVE LOGITS

nt

0.28

 potentially

0.20

 possibly

0.17

 Could

0.16

Could

0.15

could

0.14

 could

0.14

 conceivable

0.13

've

0.13

NT

0.13

Activations Density 0.058%