INDEX

Explanations

identifies functions and roles

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

ania

-0.10

ako

-0.10

 Hammond

-0.10

ank

-0.10

-0.09

 conc

-0.09

 Cain

-0.09

ovic

-0.09

 thrust

-0.09

ureau

-0.08

POSITIVE LOGITS

 play

0.17

 plays

0.16

 responsible

0.15

 involved

0.14

è´Łè´£

0.14

plays

0.13

med

0.13

 role

0.12

è§Ĵèī²

0.12

åıĤä¸İ

0.12

Activations Density 0.067%