INDEX

Explanations

states of being or condition

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

 nearing

-0.11

Dum

-0.11

 kept

-0.10

affer

-0.10

ocos

-0.10

 neat

-0.10

 being

-0.09

hij

-0.09

 slack

-0.08

POSITIVE LOGITS

 ready

0.18

rid

0.17

ready

0.14

 back

0.13

 gone

0.13

 done

0.13

 able

0.12

 airborne

0.12

.ready

0.11

 Ready

0.11

Activations Density 0.059%