APEX: Large-scale Multi-task Aesthetic-Informed Popularity Prediction for AI-Generated Music
:::info Stub — Full Engineering Breakdown Coming This paper was featured on Hugging Face Daily Papers on 2026-05-05 with 4 upvotes. A full breakdown with production viability rating, implementation notes, and honest limitations is being written. Subscribe to AI Letters → :::
| Authors | Jaavid Aktar Husain & Dorien Herremans |
| Year | 2026 |
| HF Upvotes | 4 |
| arXiv | 2605.03395 |
| Download | |
| HF Page | View on Hugging Face |
Abstract
Music popularity prediction has attracted growing research interest, with relevance to artists, platforms, and recommendation systems. However, the explosive rise of AI-generated music platforms has created an entirely new and largely unexplored landscape, where a surge of songs is produced and consumed daily without the traditional markers of artist reputation or label backing. Key, yet unexplored in this pursuit is aesthetic quality. We propose APEX, the first large-scale multi-task learning framework for AI-generated music, trained on over 211k songs (10k hours of audio) from Suno and Udio, that jointly predicts engagement-based popularity signals - streams and likes scores - alongside five perceptual aesthetic quality dimensions from frozen audio embeddings extracted from MERT, a self-supervised music understanding model. Aesthetic quality and popularity capture complementary aspects of music that together prove valuable: in an out-of-distribution evaluation on the Music Arena dataset, comprising pairwise human preference battles across eleven generative music systems unseen during training, including aesthetic features consistently improves preference prediction, demonstrating strong generalisation of the learned representations across generative architectures.
Engineering Breakdown
Plain English
This paper presents APEX, a multi-task learning framework that predicts popularity metrics (streams and likes) for AI-generated music while simultaneously scoring five dimensions of aesthetic quality. The system was trained on 211k songs (10k hours) from Suno and Udio using frozen MERT audio embeddings, establishing the first large-scale benchmark for understanding what makes AI-generated music engage listeners.
Key Engineering Insight
The core innovation is joint multi-task learning that treats aesthetic quality prediction as an auxiliary task to regularize popularity prediction—essentially using perceptual quality as an inductive bias to prevent the model from overfitting to engagement signals alone. This architectural choice directly addresses the problem that raw engagement metrics are noisy proxies for what actually makes music work.
Why It Matters for Engineers
Music platforms deploying AI-generated content need better signals than raw engagement metrics to rank and promote tracks fairly. This work solves a real production problem: distinguishing genuinely good AI music from viral novelty or platform manipulation, which matters for user retention, artist fairness, and algorithmic quality. The multi-task approach is a practical pattern engineers can apply to other domains where auxiliary quality signals improve primary task generalization.
Research Context
Prior music popularity prediction focused on traditional music with artist reputation and label signals; AI-generated music platforms lack these cues entirely, leaving a prediction gap. APEX fills this by introducing aesthetic quality as a measurable, learnable signal rather than treating it as unmeasurable. This enables the field to move from engagement-only models toward quality-aware ranking systems that can scale with the rapid growth of generative music platforms.
:::tip Subscribe Get weekly breakdowns of papers like this in AI Letters - the newsletter for engineers building production AI systems. :::