INDEX

Explanations

right or correct

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

 Left

-0.16

Left

-0.13

lhs

-0.10

_Left

-0.10

/left

-0.10

 inaccur

-0.09

.Left

-0.09

 inaccurate

-0.09

 LEFT

-0.09

rary

-0.08

POSITIVE LOGITS

 correct

0.74

 right

0.67

æŃ£ç¡®

0.49

correct

0.47

right

0.47

 ÄĳÃºng

0.46

 RIGHT

0.44

 Ð¿ÑĢÐ°Ð²Ð¸Ð»ÑĮ

0.42

 Correct

0.41

Correct

0.40

Activations Density 0.137%