INDEX

Explanations

too, enough, significant, high

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

 undes

-0.09

 Various

-0.09

igar

-0.09

Various

-0.09

 attempts

-0.09

opia

-0.09

 somewhat

-0.09

icare

-0.08

 attempt

-0.08

POSITIVE LOGITS

too

0.18

å¤ª

0.14

too

0.13

TOO

0.13

 fewer

0.13

 lots

0.13

 enough

0.12

 ÑģÐ»Ð¸ÑĪÐºÐ¾Ð¼

0.12

 pÅĻÃŃliÅ¡

0.11

å¤§éĩı

0.11

Activations Density 0.126%