INDEX

Explanations

introducing reasons

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

çĶ±äºİ

-0.14

åĽłä¸º

-0.12

 because

-0.11

/or

-0.11

because

-0.10

ï¼ĮåĽłä¸º

-0.10

ernals

-0.10

acket

-0.09

ACKET

-0.09

 thanks

-0.09

POSITIVE LOGITS

æīį

0.14

ìĿ¸ì§Ģ

0.14

ãģĵãģĿ

0.12

-ÑĤÐ¾

0.11

 importance

0.10

forward

0.10

ä¹İ

0.10

 cÃłng

0.10

why

0.10

 utmost

0.10

Activations Density 0.049%