INDEX

Explanations

willing to + verb

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

 decide

-0.10

 perfor

-0.09

åĨ³å®ļ

-0.09

 Proud

-0.09

 Hess

-0.08

nts

-0.08

 NÃ¤

-0.08

istr

-0.08

"d

-0.08

ãĥ³ãĥĨ

-0.08

POSITIVE LOGITS

 accept

0.13

bet

0.11

ermann

0.11

 wait

0.11

åĨĴ

0.11

 overlook

0.10

 ÑĥÐ³Ð¾Ð´

0.10

wait

0.10

 invest

0.10

 dialogue

0.10

Activations Density 0.031%