INDEX

Explanations

superlative qualifiers like "most" or "more"

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

ÂĢÂĢ

-0.10

¶Į

-0.10

çĮ

-0.09

 belie

-0.09

_,,

-0.08

Â¶Â¶

-0.08

/Set

-0.08

 ATTRIBUTE

-0.08

okit

-0.08

icap

-0.08

POSITIVE LOGITS

 appropriate

0.25

 correct

0.20

 fitting

0.17

 suitable

0.17

 likely

0.17

appropriate

0.15

likely

0.13

 accurate

0.13

æŃ£ç¡®

0.12

 Ð¿ÑĢÐ°Ð²Ð¸Ð»ÑĮ

0.12

Activations Density 0.026%