INDEX

Explanations

following conditional phrases

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

 Ð¸Ð¼Ñĥ

-0.15

 Ð·Ð°Ð²Ð¸ÑģÐ¸Ð¼

-0.15

 Ð½ÐµÐ´ÐµÐ»

-0.15

 Ð¿Ð¾Ð±Ð°

-0.14

 Ð´Ð¸Ð²Ð¸

-0.12

 Ð±ÑĥÐ¼Ð°

-0.12

 ÐµÑģÑĤÐµ

-0.12

 Ð·Ð°Ð±Ð¾Ð»ÐµÐ²Ð°

-0.11

 ÑħÑĢÐ°

-0.11

 Ð¾Ð³ÑĢÐ°

-0.10

POSITIVE LOGITS

âĦĸâĦĸ

0.16

 variants

0.11

 below

0.11

 variant

0.10

 following

0.10

 functionality

0.10

 nuest

0.10

 Variant

0.10

NB

0.10

 Woody

0.09

Activations Density 0.171%