Vibeleaderboard
Index / tool
Visit www.noteduel.com
Category
AI Tools
Rank
Type
TOOL
Date

About

Note Duel is a blind listening study comparing AI-composed music: listeners hear two anonymous pieces written by different models for the same compositional brief (rendered through one shared engraver and synthesizer at matched loudness) and vote for the better one, tie, or neither. Results across five briefs (minuet, invention, nocturne, march, modal miniature) are ranked with a Bradley-Terry model and 95% bootstrap confidence intervals, with per-task results published separately so no single aggregate score hides where a model actually won.

Why it made the leaderboard

Offers a rigorous, bias-controlled way to compare how different frontier models handle creative composition tasks, an area with far less established benchmarking than text generation.

Tags

ai-musicmusic-compositionllm-arenablind-testranking

Comments (0)

No comments yet

Editorially curated, with community endorsements as a secondary signal. Corrections welcome.