Call for Papers
The LLM4MA workshop invites submissions of research and position papers that present early-stage ideas, empirical findings, or visionary perspectives at the intersection of large language models and music/audio. All submissions will be peer-reviewed under a double-blind review process, meaning that authors must anonymize their manuscripts by removing all identifying information. Submissions should follow the
Paper Template for ISMIR 2025 LLM4Music Satellite Event.zip (adapted from the ISMIR 2025 style).
- Extended Abstracts: Up to 2 pages (excluding references and appendix); suitable for early-stage research, conceptual explorations, or in-progress work.
- Short Papers: Up to 4 pages (excluding references and appendix); suitable for more developed findings or implementation-focused contributions.
We welcome contributions on topics including, but not limited to:
- Tokenization of Music: Novel encoding schemes for symbolic or audio representations compatible with LLMs.
- Long Sequence Modeling: Advancing architectures and training techniques for generating and understanding extended musical compositions.
- Instruction and Reinforcement Fine-Tuning: Aligning LLMs with creative music instructions, tasks, and feedback signals.
- Artistic Creation and Evaluation: Methods for enhancing creativity, measuring aesthetic value, and evaluating originality in AI-generated music.
- Industry Applications: Use cases of music/audio LLMs in film scoring, game soundtracks, music recommendation, and other commercial contexts.
- Enhanced Musical Diversity: Techniques to support culturally diverse, stylistically rich, and inclusive music generation.
- Scaling Laws and Emergent Abilities: How model scaling affects musical capabilities and emergent musical behaviors.
- Dataset Design and Generalizability: Strategies for building multilingual, multi-style datasets and evaluating model robustness across musical domains.
- Modality Extension: Enabling models to process both acoustic waveforms and symbolic music data for tasks like transcription, composition, or control.
- Pre-training Paradigms: Exploration of self-supervised, multi-task, or multi-modal pre-training for music/audio models.
- Inference and Prompt Engineering: Prompt-based control and output optimization for musical creativity and task accuracy.
- In-context Learning and Chain-of-Thought: Leveraging LLM reasoning and few-shot learning in musical generation or analysis tasks.
- Autonomous Music Agents: Systems that can independently compose, interact, and adapt to musical goals or user feedback.
Accepted papers will be presented as posters, with selected contributions featured in oral spotlight sessions. A small number of submissions may be highlighted for special mentions or invited for further discussion in our panel sessions.
Poster Presentation Options
We offer flexible presentation options for accepted papers:
- In-person Presentation: Present your poster at the venue during the poster session.
- Virtual Presentation: For authors unable to attend in person, we welcome virtual poster presentations. Please prepare a 5-minute video presentation of your work to be shared during the poster session.
- Interactive Online Q&A: Authors can include a QR code on their poster linking to a dedicated Zoom room for real-time Q&A sessions with attendees.
Note: We will help with poster printing for both in-person and virtual presentations. Please contact the organizers for printing arrangements.
Novelty and Openness
We accept submissions that are under review elsewhere or intended for future conference submission. However, submissions must not have been formally published at other venues. We strongly encourage open science practices — including code, datasets, checkpoints, and training pipelines — to enhance transparency and reproducibility. Accepted papers will be published on the website, but the workshop is non-archival.
Key Dates
- Submission Open: 6th June, 2025
- Submission Deadline:
10th August, 2025 17th August, 2025
- Notifications: 31st August, 2025
- Camera-Ready: 14th September, 2025
- Workshop: 26th September, 2025
