INDEX

Explanations

demonstrate or explain

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

 drawing

-0.11

-0.10

ola

-0.10

 Detailed

-0.09

 Drawing

-0.09

 Expedition

-0.09

 rendering

-0.09

 rever

-0.08

 stating

-0.08

wor

-0.08

POSITIVE LOGITS

 demon

0.35

 Ð´ÐµÐ¼Ð¾Ð½

0.33

 demonstration

0.33

 demonstrate

0.32

 Demon

0.31

Demon

0.30

ç¤º

0.30

 show

0.27

å±ķç¤º

0.26

 demonstrates

0.25

Activations Density 0.248%