INDEX

Explanations

strategies and advice for action

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

enie

-0.10

sop

-0.09

pong

-0.09

 iceberg

-0.09

exampleInputEmail

-0.09

enia

-0.09

azer

-0.08

ine

-0.08

enko

-0.08

 Sheldon

-0.08

POSITIVE LOGITS

how

0.19

å¦Ĥä½ķ

0.19

how

0.17

 ÚĨÚ¯ÙĪÙĨÙĩ

0.13

 cÃ³mo

0.13

 hvordan

0.12

 ways

0.12

 ÙĥÙĬÙģ

0.12

 tips

0.12

How

0.12

Activations Density 0.140%