ASTAR-NTU solution to AudioMOS Challenge 2025 Track1

2025/07/14 by Ritter-Gutierrez, Fabian, Lin, Yi-Cheng, Wei, Jui-Chiang +3
#Audio and Speech Processing (eess.AS) #FOS: Computer and information sciences #FOS: Electrical engineering #Sound (cs.SD) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2507.09904

Abstract

Evaluation of text-to-music systems is constrained by the cost and availability of collecting experts for assessment. AudioMOS 2025 Challenge track 1 is created to automatically predict music impression (MI) as well as text alignment (TA) between the prompt and the generated musical piece. This paper reports our winning system, which uses a dual-branch architecture with pre-trained MuQ and RoBERTa models as audio and text encoders. A cross-attention mechanism fuses the audio and text representations. For training, we reframe the MI and TA prediction as a classification task. To incorporate the ordinal nature of MOS scores, one-hot labels are converted to a soft distribution using a Gaussian kernel. On the official test set, a single model trained with this method achieves a system-level Spearman's Rank Correlation Coefficient (SRCC) of 0.991 for MI and 0.952 for TA, corresponding to a relative improvement of 21.21% in MI SRCC and 31.47% in TA SRCC over the challenge baseline.

Citations

• Meta-PerSER: Few-Shot Listener Personalized Speech Emotion Recognition via Meta-learning
• MuQ: Self-Supervised Music Representation Learning with Mel Residual Vector Quantization
• Mamba: Linear-Time Sequence Modeling with Selective State Spaces
• AudioLDM 2: Learning Holistic Audio Generation with Self-supervised Pretraining
• Simple and Controllable Music Generation
• MuseCoco: Generating Symbolic Music from Text
• ERNIE-Music: Text-to-Waveform Music Generation with Diffusion Models
• Noise2Music: Text-conditioned Music Generation with Diffusion Models
• Make-An-Audio: Text-To-Audio Generation with Prompt-Enhanced Diffusion Models
• AudioLDM: Text-to-Audio Generation with Latent Diffusion Models
• Moûsai: Text-to-Music Generation with Long-Context Latent Diffusion
• MuChoMusic dataset
• Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation
• AudioGen: Textually Guided Audio Generation
• RoBERTa: A Robustly Optimized BERT Pretraining Approach
• Rank consistent ordinal regression for neural networks with application to age estimation
• Attention Is All You Need
• Framewise phoneme classification with bidirectional LSTM and other neural network architectures
• Ridge Regression: Biased Estimation for Nonorthogonal Problems
• MusicEval: A Generative Music Dataset with Expert Ratings for Automatic Text-to-Music Evaluation

Related

• WhisQ: Cross-Modal Representation Learning for Text-to-Music MOS Prediction
• QAMRO: Quality-aware Adaptive Margin Ranking Optimization for Human-aligned Assessment of Audio Generation Systems
• The AudioMOS Challenge 2025
• The T12 System for AudioMOS Challenge 2025: Audio Aesthetics Score Prediction System Using KAN- and VERSA-based Models
• The MUSE Benchmark: Probing Music Perception and Auditory Relational Reasoning in Audio LLMS