INDEX

Explanations

the followed by specific points/positions

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

 pinnacle

-0.13

 onset

-0.12

 outset

-0.12

 zenith

-0.10

 apex

-0.10

 inception

-0.09

 midst

-0.09

 entirety

-0.09

 Apex

-0.09

illard

-0.09

POSITIVE LOGITS

 expense

0.25

 discretion

0.19

end

0.18

 moment

0.17

expense

0.16

 same

0.16

 Expense

0.16

 mercy

0.15

 request

0.14

 seams

0.13

Activations Density 0.035%