INDEX

Explanations

groups and their descriptions

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

 both

-0.12

 BOTH

-0.12

 beide

-0.12

 ambos

-0.12

ä¸¡

-0.11

 Both

-0.11

 twin

-0.11

 Ð¾Ð±Ð¾

-0.11

 beiden

-0.10

ä¸¤äºº

-0.10

POSITIVE LOGITS

0.18

 Ø¹Ø¨Ø§Ø±Øª

0.17

åĪĨåĪ«

0.16

?,?,

0.13

à¹īà¹ģà¸ģ

0.12

 Bunlar

0.11

 à¹Ħà¸Ķ

0.11

 including

0.11

_OCCURRED

0.11

à¹Į:

0.11

Activations Density 0.267%