INDEX

Explanations

everything changed

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

-0.12

chw

-0.09

.toolbox

-0.09

arrass

-0.09

 mirac

-0.09

 Coff

-0.08

ouch

-0.08

åħ¸

-0.08

 Gardner

-0.08

 paradox

-0.08

POSITIVE LOGITS

 everything

0.28

everything

0.22

 Everything

0.21

Everything

0.20

ä¸ĢåĪĩ

0.18

 changed

0.18

 tudo

0.17

 alles

0.15

 life

0.14

 lives

0.14

Activations Density 0.117%