INDEX

Explanations

functioning, interaction, and behavior

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

 Ð·Ð°ÑıÐ²

-0.15

 ÑĦÑĥÐ½Ð´Ð°

-0.13

 Ð¸Ð¼Ñĥ

-0.13

 Ð¸Ð½Ð´Ð¸Ð²Ð¸Ð´Ñĥ

-0.12

 Ð¾Ð³ÑĢÐ°

-0.12

 Ð¼ÐµÑĤÐ°Ð»Ð»Ð¸

-0.12

 ÑĤÐ°Ð±Ð»Ð¸

-0.11

 Ð¿Ð¾Ð±Ð°

-0.10

 Ð´Ð¾ÐºÑĥÐ¼

-0.10

 Crud

-0.10

POSITIVE LOGITS

 functioning

0.19

 processes

0.13

 process

0.13

 structure

0.12

process

0.11

 activity

0.11

zier

0.11

 relations

0.11

 hoáº¡t

0.10

 entirety

0.10

Activations Density 0.111%