INDEX

Explanations

as follows: is are of

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

anko

-0.10

ullo

-0.10

 incor

-0.09

ano

-0.09

oint

-0.09

fam

-0.09

lar

-0.09

 Reco

-0.08

cer

-0.08

vor

-0.08

POSITIVE LOGITS

å¦Ĥä¸ĭ

0.33

 follows

0.20

 reads

0.18

 Ð²ÑĭÐ³Ð»ÑıÐ´

0.18

 looks

0.17

 vypad

0.16

 Reads

0.15

reads

0.14

 Ø¹Ø¨Ø§Ø±Øª

0.14

 ëĭ¤ìĿĮê³¼

0.13

Activations Density 0.158%