INDEX

Explanations

sorry

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

 correct

-0.09

SM

-0.09

GenerationStrategy

-0.09

 Wend

-0.09

TRA

-0.09

 means

-0.09

 indeed

-0.09

á»ĳng

-0.09

ÑĪÐ¸Ð±

-0.09

ï½ŀï½ŀ

-0.09

POSITIVE LOGITS

 sorry

0.24

 Sorry

0.22

Sorry

0.22

sorry

0.22

SOR

0.17

æĬ±

0.16

orry

0.16

ä¸įèĥ½

0.13

 lament

0.13

Sor

0.13

Activations Density 0.135%