INDEX

Explanations

how and ability

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

utin

-0.11

 Precision

-0.10

iggins

-0.09

ordum

-0.09

onis

-0.09

 Lambert

-0.09

-0.08

ker

-0.08

Precision

-0.08

POSITIVE LOGITS

how

0.18

 ability

0.14

how

0.12

å¦Ĥä½ķ

0.12

 enough

0.12

èĥ½å¤Ł

0.11

 able

0.11

 appreciation

0.10

 Ability

0.10

 cÃ³mo

0.10

Activations Density 0.054%