Music Arena: Live Evaluation for Text-to-Music
Yonghyun Kim, Wayne Chi, Anastasios N. Angelopoulos, Wei-Lin Chiang, Koichi Saito, Shinji Watanabe, Yuki Mitsufuji, Chris Donahue
2025-07-29
Summary
This paper talks about Music Arena, an online platform where people can listen to music created by AI from text descriptions and give their opinions in real time.
What's the problem?
The problem is that evaluating AI-generated music is hard because different people have different tastes, and there is no standard way to collect feedback quickly and fairly. Also, privacy concerns make it difficult to collect honest opinions.
What's the solution?
Music Arena solves this by providing a standard way for people to evaluate music created from text prompts, allowing detailed feedback while protecting users' privacy. This makes the evaluation process open, fair, and efficient.
Why it matters?
This matters because it helps researchers and developers understand how well their AI models create music that people actually enjoy, leading to better and more creative music-making tools in the future.
Abstract
Music Arena is an open platform for real-time human preference evaluation of text-to-music models, featuring a standardized protocol, detailed feedback, and privacy guarantees.