INDEX

Explanations

adding "additionally"

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

 indeed

-0.12

ÌĨ

-0.09

 jednak

-0.09

äºİæĺ¯

-0.09

 totiÅ¾

-0.09

à¸Ńà¸Ķ

-0.08

powered

-0.08

 Whip

-0.08

Ð°Ð

-0.08

 Ð²Ð´ÑĢÑĥÐ³

-0.08

POSITIVE LOGITS

 since

0.14

note

0.13

 because

0.13

 importantly

0.13

 note

0.12

 Ð¿Ð¾ÑģÐºÐ¾Ð»ÑĮÐºÑĥ

0.12

çĶ±äºİ

0.12

 there

0.11

there

0.11

 ÙģØ¥ÙĨ

0.11

Activations Density 0.030%