Pay (Cross) Attention to the Melody: Curriculum Masking for Single-Encoder Melodic Harmonization

Melodic harmonization, the task of generating harmonic accompaniments for a given melody, remains a central challenge in computational music generation. Recent single encoder transformer approaches have framed harmonization as a masked sequence modeling problem, but existing training curricula inspired by discrete diffusion often result in weak (cross) attention between melody and harmony. This leads to limited exploitation of melodic cues, particularly in out-of-domain contexts. In this work, we introduce a training curriculum, FF (full-to-full), which keeps all harmony tokens masked for several training steps before progressively unmasking entire sequences during training to strengthen melody-harmony interactions. We systematically evaluate this approach against prior curricula across multiple experimental axes, including temporal quantization (quarter vs. sixteenth note), bar-level vs. time-signature conditioning, melody representation (full range vs. pitch class), and inference-time unmasking strategies. Models are trained on the HookTheory dataset and evaluated both in-domain and on a curated collection of jazz standards, using a comprehensive set of metrics that assess chord progression structure, harmony-melody alignment, and rhythmic coherence. Results demonstrate that the proposed FF curriculum consistently outperforms baselines in nearly all metrics, with particularly strong gains in out-of-domain evaluations where harmonic adaptability to novel melodic queues is crucial. We further find that quarter-note quantization, intertwining of bar tokens, and pitch-class melody representations are advantageous in the FF setting. Our findings highlight the importance of training curricula in enabling effective melody conditioning and suggest that full-to-full unmasking offers a robust strategy for single encoder harmonization.

Divergence-Free Diffusion Models for Incompressible Fluid Flows

Generative diffusion models are extensively used in unsupervised and self-supervised machine learning with the aim to generate new samples from a probability distribution estimated with a set of known samples. They have demonstrated impressive results in replicating dense, real-world contents such as images, musical pieces, or human languages. This work investigates their application to the numerical simulation of incompressible fluid flows, with a view toward incorporating physical constraints such as incompressibility in the probabilistic forecasting framework enabled by generative networks. For that purpose, we explore different conditional, score-based diffusion models where the divergence-free constraint is imposed by the Leray spectral projector, and autoregressive conditioning is aimed at stabilizing the forecasted flow snapshots at distant time horizons. The proposed models are run on a benchmark turbulence problem, namely a Kolmogorov flow, which allows for a fairly detailed analytical and numerical treatment and thus simplifies the evaluation of the numerical methods used to simulate it. Numerical experiments of increasing complexity are performed in order to compare the advantages and limitations of the diffusion models we have implemented and appraise their performances, including: (i) in-distribution assessment over the same time horizons and for similar physical conditions as the ones seen during training; (ii) rollout predictions over time horizons unseen during training; and (iii) out-of-distribution tests for forecasting flows markedly different from those seen during training. In particular, these results illustrate the ability of diffusion models to reproduce the main statistical characteristics of Kolmogorov turbulence in scenarios departing from the ones they were trained on.

Song Aesthetics Evaluation with Multi-Stem Attention and Hierarchical Uncertainty Modeling

Music generative artificial intelligence (AI) is rapidly expanding music content, necessitating automated song aesthetics evaluation. However, existing studies largely focus on speech, audio or singing quality, leaving song aesthetics underexplored. Moreover, conventional approaches often predict a precise Mean Opinion Score (MOS) value directly, which struggles to capture the nuances of human perception in song aesthetics evaluation. This paper proposes a song-oriented aesthetics evaluation framework, featuring two novel modules: 1) Multi-Stem Attention Fusion (MSAF) builds bidirectional cross-attention between mixture-vocal and mixture-accompaniment pairs, fusing them to capture complex musical features; 2) Hierarchical Granularity-Aware Interval Aggregation (HiGIA) learns multi-granularity score probability distributions, aggregates them into a score interval, and applies a regression within the interval to produce the final score. We evaluated on two datasets of full-length songs: SongEval dataset (AI-generated) and an internal aesthetics dataset (human-created), and compared with two state-of-the-art (SOTA) models. Results show that the proposed method achieves stronger performance for multi-dimensional song aesthetics evaluation.

FusID: Modality-Fused Semantic IDs for Generative Music Recommendation

Generative recommendation systems have achieved significant advances by leveraging semantic IDs to represent items. However, existing approaches that tokenize each modality independently face two critical limitations: (1) redundancy across modalities that reduces efficiency, and (2) failure to capture inter-modal interactions that limits item representation. We introduce FusID, a modality-fused semantic ID framework that addresses these limitations through three key components: (i) multimodal fusion that learns unified representations by jointly encoding information across modalities, (ii) representation learning that brings frequently co-occurring item embeddings closer while maintaining distinctiveness and preventing feature redundancy, and (iii) product quantization that converts the fused continuous embeddings into multiple discrete tokens to mitigate ID conflict. Evaluated on a multimodal next-song recommendation (i.e., playlist continuation) benchmark, FusID achieves zero ID conflicts, ensuring that each token sequence maps to exactly one song, mitigates codebook underutilization, and outperforms baselines in terms of MRR and Recall@k (k = 1, 5, 10, 20).

Inverting Self-Organizing Maps: A Unified Activation-Based Framework

Self-Organizing Maps provide topology-preserving projections of high-dimensional data and have been widely used for visualization, clustering, and vector quantization. In this work, we show that the activation pattern of a SOM - the squared distances to its prototypes - can be inverted to recover the exact input under mild geometric conditions. This follows from a classical fact in Euclidean distance geometry: a point in $D$ dimensions is uniquely determined by its distances to $D{+}1$ affinely independent references. We derive the corresponding linear system and characterize the conditions under which the inversion is well-posed. Building upon this mechanism, we introduce the Manifold-Aware Unified SOM Inversion and Control (MUSIC) update rule, which enables controlled, semantically meaningful trajectories in latent space. MUSIC modifies squared distances to selected prototypes while preserving others, resulting in a deterministic geometric flow aligned with the SOM's piecewise-linear structure. Tikhonov regularization stabilizes the update rule and ensures smooth motion on high-dimensional datasets. Unlike variational or probabilistic generative models, MUSIC does not rely on sampling, latent priors, or encoder-decoder architectures. If no perturbation is applied, inversion recovers the exact input; when a target cluster or prototype is specified, MUSIC produces coherent semantic variations while remaining on the data manifold. This leads to a new perspective on data augmentation and controllable latent exploration based solely on prototype geometry. We validate the approach using synthetic Gaussian mixtures, the MNIST and the Faces in the Wild dataset. Across all settings, MUSIC produces smooth, interpretable trajectories that reveal the underlying geometry of the learned manifold, illustrating the advantages of SOM-based inversion over unsupervised clustering.

Mathematical Foundations of Polyphonic Music Generation via Structural Inductive Bias

This monograph introduces a novel approach to polyphonic music generation by addressing the "Missing Middle" problem through structural inductive bias. Focusing on Beethoven's piano sonatas as a case study, we empirically verify the independence of pitch and hand attributes using normalized mutual information (NMI=0.167) and propose the Smart Embedding architecture, achieving a 48.30% reduction in parameters. We provide rigorous mathematical proofs using information theory (negligible loss bounded at 0.153 bits), Rademacher complexity (28.09% tighter generalization bound), and category theory to demonstrate improved stability and generalization. Empirical results show a 9.47% reduction in validation loss, confirmed by SVD analysis and an expert listening study (N=53). This dual theoretical and applied framework bridges gaps in AI music generation, offering verifiable insights for mathematically grounded deep learning.

The ICASSP 2026 Automatic Song Aesthetics Evaluation Challenge

This paper summarizes the ICASSP 2026 Automatic Song Aesthetics Evaluation (ASAE) Challenge, which focuses on predicting the subjective aesthetic scores of AI-generated songs. The challenge consists of two tracks: Track 1 targets the prediction of the overall musicality score, while Track 2 focuses on predicting five fine-grained aesthetic scores. The challenge attracted strong interest from the research community and received numerous submissions from both academia and industry. Top-performing systems significantly surpassed the official baseline, demonstrating substantial progress in aligning objective metrics with human aesthetic preferences. The outcomes establish a standardized benchmark and advance human-aligned evaluation methodologies for modern music generation systems.

VoxMorph: Scalable Zero-shot Voice Identity Morphing via Disentangled Embeddings

Morphing techniques generate artificial biometric samples that combine features from multiple individuals, allowing each contributor to be verified against a single enrolled template. While extensively studied in face recognition, this vulnerability remains largely unexplored in voice biometrics. Prior work on voice morphing is computationally expensive, non-scalable, and limited to acoustically similar identity pairs, constraining practical deployment. Moreover, existing sound-morphing methods target audio textures, music, or environmental sounds and are not transferable to voice identity manipulation. We propose VoxMorph, a zero-shot framework that produces high-fidelity voice morphs from as little as five seconds of audio per subject without model retraining. Our method disentangles vocal traits into prosody and timbre embeddings, enabling fine-grained interpolation of speaking style and identity. These embeddings are fused via Spherical Linear Interpolation (Slerp) and synthesized using an autoregressive language model coupled with a Conditional Flow Matching network. VoxMorph achieves state-of-the-art performance, delivering a 2.6x gain in audio quality, a 73% reduction in intelligibility errors, and a 67.8% morphing attack success rate on automated speaker verification systems under strict security thresholds. This work establishes a practical and scalable paradigm for voice morphing with significant implications for biometric security. The code and dataset are available on our project page: https://vcbsl.github.io/VoxMorph/

Abusive music and song transformation using GenAI and LLMs

Repeated exposure to violence and abusive content in music and song content can influence listeners' emotions and behaviours, potentially normalising aggression or reinforcing harmful stereotypes. In this study, we explore the use of generative artificial intelligence (GenAI) and Large Language Models (LLMs) to automatically transform abusive words (vocal delivery) and lyrical content in popular music. Rather than simply muting or replacing a single word, our approach transforms the tone, intensity, and sentiment, thus not altering just the lyrics, but how it is expressed. We present a comparative analysis of four selected English songs and their transformed counterparts, evaluating changes through both acoustic and sentiment-based lenses. Our findings indicate that Gen-AI significantly reduces vocal aggressiveness, with acoustic analysis showing improvements in Harmonic to Noise Ratio, Cepstral Peak Prominence, and Shimmer. Sentiment analysis reduced aggression by 63.3-85.6\% across artists, with major improvements in chorus sections (up to 88.6\% reduction). The transformed versions maintained musical coherence while mitigating harmful content, offering a promising alternative to traditional content moderation that avoids triggering the "forbidden fruit" effect, where the censored content becomes more appealing simply because it is restricted. This approach demonstrates the potential for GenAI to create safer listening experiences while preserving artistic expression.

ConceptCaps -- a Distilled Concept Dataset for Interpretability in Music Models

Concept-based interpretability methods like TCAV require clean, well-separated positive and negative examples for each concept. Existing music datasets lack this structure: tags are sparse, noisy, or ill-defined. We introduce ConceptCaps, a dataset of 23k music-caption-audio triplets with explicit labels from a 200-attribute taxonomy. Our pipeline separates semantic modeling from text generation: a VAE learns plausible attribute co-occurrence patterns, a fine-tuned LLM converts attribute lists into professional descriptions, and MusicGen synthesizes corresponding audio. This separation improves coherence and controllability over end-to-end approaches. We validate the dataset through audio-text alignment (CLAP), linguistic quality metrics (BERTScore, MAUVE), and TCAV analysis confirming that concept probes recover musically meaningful patterns. Dataset and code are available online.