INDEX

Explanations

develop plans and think

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

ì§ĵ

-0.10

 Higgins

-0.10

 Morgan

-0.09

 Foley

-0.09

ohn

-0.09

 trif

-0.09

Sor

-0.09

 Boone

-0.09

Slinky

-0.08

 trailing

-0.08

POSITIVE LOGITS

 horizon

0.12

 business

0.12

 Recomm

0.12

 recommend

0.11

 recommending

0.11

 colleagues

0.10

-translate

0.10

 department

0.10

 recommendations

0.10

recommend

0.10

Activations Density 0.076%