INDEX

Explanations

could be a

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

amt

-0.10

anner

-0.09

 inevitable

-0.08

CEE

-0.08

enes

-0.08

atile

-0.08

lander

-0.08

èĩ³å°ĳ

-0.08

annah

-0.08

 Richt

-0.08

POSITIVE LOGITS

 anything

0.31

anything

0.23

 Anything

0.21

Anything

0.20

 anywhere

0.15

 something

0.14

 everything

0.14

 either

0.14

 simple

0.14

any

0.13

Activations Density 0.139%