Selected work

Representative Papers

A Mamba-Based Model for Automatic Chord Recognition

Jan 05, 2026arXiv.org

This work addresses the trade-off between efficiency and accuracy in modeling long-range temporal dependencies for automatic chord recognition by proposing BMACE, the first model to introduce a bidirectional Mamba architecture to this task. Built upon selective structured state space models, BMACE integrates bidirectional temporal modeling with deep learning–based audio feature extraction to effectively capture long-range dependencies in musical signals. Experimental results demonstrate that BMACE achieves chord recognition accuracy comparable to state-of-the-art methods on standard benchmarks while substantially reducing both model parameters and computational overhead, thereby unifying high accuracy with high efficiency.

0 citationsRead paper

BeatlesFC: Harmonic function annotations of Isophonics' The Beatles dataset

Jan 05, 2026arXiv.org

This study addresses the lack of structured harmonic function annotations—such as tonic, dominant, and subdominant—in existing Beatles song datasets, which has hindered the integration of chord progressions with macro-level formal structure. To bridge this gap, the work introduces harmonic function theory systematically into a popular music dataset for the first time, constructing a binary annotation scheme that distinguishes stable (tonic) from unstable (subdominant and dominant) functions grounded in music theory. By combining expert annotations with structural parsing, the research establishes connections between chords and musical form at the phrase level. The resulting resource, BeatlesFC, extends the Isophonics dataset with comprehensive harmonic function labels spanning all Beatles albums, thereby filling a critical gap between chord-level and form-level analysis and offering a valuable new asset for music information retrieval and automated music analysis.

0 citationsRead paper

Story2MIDI: Emotionally Aligned Music Generation from Text

Dec 01, 2025

This work addresses the problem of generating emotion-consistent music from text. We propose a cross-modal sequence-to-sequence Transformer framework, whose core innovations are: (1) constructing the first small-scale paired dataset integrating fine-grained textual emotion annotations with corresponding musical affective representations (e.g., valence and arousal), and (2) designing an emotion-aligned attention mechanism that explicitly models semantic–affective mappings between text and music. The model is rigorously evaluated via objective musical metrics—including tonal coherence and rhythmic stability—as well as double-blind human listening experiments. Results demonstrate statistically significant improvements in both emotional alignment (p < 0.01) and perceptual naturalness, while preserving narrative coherence. This work establishes a novel, interpretable, and empirically evaluable paradigm for emotion-driven AI composition.

0 citationsRead paper

Chord-conditioned Melody and Bass Generation

Nov 11, 2025

This study addresses joint melody and bassline generation conditioned on chord progressions, aiming to improve theoretical correctness and stylistic coherence in pitch content, interval distributions, and chord-tone usage. We propose five Transformer-based chord-conditioned generation paradigms and conduct a systematic comparative analysis of their modeling capabilities. Crucially, we introduce the first “bass-first” two-stage generation strategy, wherein bassline generation is performed prior to melody synthesis and serves as a strong structural constraint. Leveraging music-theory-driven quantitative metrics—including chord-tone coverage and interval distribution deviation—we empirically demonstrate that chord conditioning significantly enhances generation quality. The bass-first model achieves superior stylistic fidelity, particularly in tonal logic and voice-leading plausibility, outperforming baselines by an average of 12.7%. This work establishes a novel paradigm for incorporating structured musical priors into generative models.

0 citationsRead paper

Picat Through the Lens of Advent of Code

Jul 15, 2025

This study evaluates the applicability and efficiency of the Picat programming language for solving algorithmic problems involving reverse engineering and pathfinding, using multiple challenges from Advent of Code (AoC) 2024 as benchmarks. To address such problems, we propose a declarative modeling approach that integrates SAT-based constraint solving with tabling—a memoization mechanism for dynamic programming—leveraging Picat’s native support for logic programming, constraint reasoning, pattern matching, and tabled recursion. Our method eliminates manual optimization while preserving clarity and conciseness, reducing code size by 30–60%. Empirically, it achieves speedups of up to one to two orders of magnitude on several benchmarks. Experimental results demonstrate that Picat delivers both high expressive power and strong runtime performance for complex search and combinatorial reasoning tasks, confirming its viability as an efficient tool for algorithm prototyping and competitive programming.

0 citationsRead paper

Recent publications

Latest Papers

A Mamba-Based Model for Automatic Chord Recognition

Jan 05, 2026arXiv.org

This work addresses the trade-off between efficiency and accuracy in modeling long-range temporal dependencies for automatic chord recognition by proposing BMACE, the first model to introduce a bidirectional Mamba architecture to this task. Built upon selective structured state space models, BMACE integrates bidirectional temporal modeling with deep learning–based audio feature extraction to effectively capture long-range dependencies in musical signals. Experimental results demonstrate that BMACE achieves chord recognition accuracy comparable to state-of-the-art methods on standard benchmarks while substantially reducing both model parameters and computational overhead, thereby unifying high accuracy with high efficiency.

0 citationsRead paper

BeatlesFC: Harmonic function annotations of Isophonics' The Beatles dataset

Jan 05, 2026arXiv.org

This study addresses the lack of structured harmonic function annotations—such as tonic, dominant, and subdominant—in existing Beatles song datasets, which has hindered the integration of chord progressions with macro-level formal structure. To bridge this gap, the work introduces harmonic function theory systematically into a popular music dataset for the first time, constructing a binary annotation scheme that distinguishes stable (tonic) from unstable (subdominant and dominant) functions grounded in music theory. By combining expert annotations with structural parsing, the research establishes connections between chords and musical form at the phrase level. The resulting resource, BeatlesFC, extends the Isophonics dataset with comprehensive harmonic function labels spanning all Beatles albums, thereby filling a critical gap between chord-level and form-level analysis and offering a valuable new asset for music information retrieval and automated music analysis.

0 citationsRead paper

Story2MIDI: Emotionally Aligned Music Generation from Text

Dec 01, 2025

This work addresses the problem of generating emotion-consistent music from text. We propose a cross-modal sequence-to-sequence Transformer framework, whose core innovations are: (1) constructing the first small-scale paired dataset integrating fine-grained textual emotion annotations with corresponding musical affective representations (e.g., valence and arousal), and (2) designing an emotion-aligned attention mechanism that explicitly models semantic–affective mappings between text and music. The model is rigorously evaluated via objective musical metrics—including tonal coherence and rhythmic stability—as well as double-blind human listening experiments. Results demonstrate statistically significant improvements in both emotional alignment (p < 0.01) and perceptual naturalness, while preserving narrative coherence. This work establishes a novel, interpretable, and empirically evaluable paradigm for emotion-driven AI composition.

0 citationsRead paper

Chord-conditioned Melody and Bass Generation

Nov 11, 2025

This study addresses joint melody and bassline generation conditioned on chord progressions, aiming to improve theoretical correctness and stylistic coherence in pitch content, interval distributions, and chord-tone usage. We propose five Transformer-based chord-conditioned generation paradigms and conduct a systematic comparative analysis of their modeling capabilities. Crucially, we introduce the first “bass-first” two-stage generation strategy, wherein bassline generation is performed prior to melody synthesis and serves as a strong structural constraint. Leveraging music-theory-driven quantitative metrics—including chord-tone coverage and interval distribution deviation—we empirically demonstrate that chord conditioning significantly enhances generation quality. The bass-first model achieves superior stylistic fidelity, particularly in tonal logic and voice-leading plausibility, outperforming baselines by an average of 12.7%. This work establishes a novel paradigm for incorporating structured musical priors into generative models.

0 citationsRead paper

Picat Through the Lens of Advent of Code

Jul 15, 2025

This study evaluates the applicability and efficiency of the Picat programming language for solving algorithmic problems involving reverse engineering and pathfinding, using multiple challenges from Advent of Code (AoC) 2024 as benchmarks. To address such problems, we propose a declarative modeling approach that integrates SAT-based constraint solving with tabling—a memoization mechanism for dynamic programming—leveraging Picat’s native support for logic programming, constraint reasoning, pattern matching, and tabled recursion. Our method eliminates manual optimization while preserving clarity and conciseness, reducing code size by 30–60%. Empirically, it achieves speedups of up to one to two orders of magnitude on several benchmarks. Experimental results demonstrate that Picat delivers both high expressive power and strong runtime performance for complex search and combinatorial reasoning tasks, confirming its viability as an efficient tool for algorithm prototyping and competitive programming.

0 citationsRead paper