By Oriol Nieto, Senior Research Engineer II, and Gautham Mysore, Senior Principal Scientist
Adobe Research took a leading role in organizing the International Society for Music Information Retrieval Conference (ISMIR), the premiere conference in music AI. For the 25th anniversary of ISMIR in 2024, we joined forces with Blair Kaneshiro (adjunct professor at Stanford University and University of Alaska, Anchorage) as the conference’s general chairs. Together, we envisioned, designed, and implemented the event, which took place in San Francisco, California, from November 10 to 15, 2024. Here, we share some highlights and reflections on this monumental event.
A bridging theme
Our theme for ISMIR 2024 was “Bridging Technology and Musical Creativity.” This theme was particularly fitting given our location in San Francisco, a city renowned for its iconic bridges and innovative spirit. The theme underscored the conference’s focus on connecting the worlds of technology and music as well as fostering collaboration and creativity. To honor this theme, we organized several creative practices events, featuring leaders in music technology and acclaimed composers/artists who helped shape the conversation and connect these two often disjointed communities.
Wealth of research
ISMIR 2024 published a total of 123 papers, showcasing cutting-edge research in the field of music information retrieval, including some of the latest advances in AI for music generation. In addition, the conference included 6 accepted tutorials presented during the first day. The range and depth of the research presented were truly inspiring.
Submissions, acceptance rate, and open reviews
We received a record high number of submissions for this conference, and the acceptance rate was the lowest ever for ISMIR (35.55 percent). This reflects the high quality of submissions we received and the rigorous review process that ensures only the most innovative research is presented at ISMIR. Additionally, we had a record number of tutorial submissions.
For the first time at ISMIR, we publicly released a subset of the paper reviews so that the community could read what their peers think of some of the published works. Authors, reviewers, and meta–reviewers who gave their consent had their reviews published alongside their accepted papers. This initiative aimed to increase transparency and foster a culture of constructive feedback within the community.
Record attendance and hybrid format
We were thrilled to welcome a record number of in-person attendees, with more than 530 participants joining us in San Francisco. Around 60 percent came from abroad. The hybrid format allowed us to extend our reach even further, with over 200 virtual attendees participating from around the globe. Our virtual program ran seamlessly thanks to the dedication and hard work of our organizing team, enabling us to provide a rich and engaging experience for those who couldn’t join us in person. This inclusive approach ensured that everyone, regardless of location, could be part of the ISMIR community.
Emphasizing accessibility
For the first time in ISMIR’s history, we appointed an accessibility chair. This role was crucial in making the conference more inclusive and accommodating. From providing COVID-19 tests to ensuring that PDFs were compatible with colorblindness, we took significant steps to provide for the diverse needs of our attendees.
Banquet and jam session
Because this was a music-focused conference, we wanted to celebrate with some tunes. One of the highlights of the conference was the banquet and a performance by Wil Blades, a California-based jazz fusion artist, along with an on-stage jam session that brought together musicians and researchers in a celebration of creativity and collaboration in the legendary Regency Ballroom in the heart of San Francisco.
Adobe Research’s contributions
Adobe Researchers have contributed groundbreaking work to ISMIR for over a decade. At ISMIR 2024, we published two papers that yield state-of-the art results in the areas of large language music models and music generation, respectively:
- “Augment, Drop & Swap: Improving Diversity in LLM Captions for Efficient Music-Text Representation Learning” by Ilaria Manco, Justin Salamon, and Oriol Nieto
- “DITTO-2: Distilled Diffusion Inference-Time T-Optimization for Music Generation” by Zachary Novack, Julian McAuley, Taylor Berg-Kirkpatrick, and Nicholas J. Bryan
Adobe researchers and interns also presented a world-class tutorial:
- “Connecting Music Audio and Natural Language” by Seung Heon Doh, Ilaria Manco, Zachary Novack, Jong Wook Kim, and Ke Chen
In addition, Adobe was one of the Platinum Sponsors at this year’s ISMIR, which was crucial for the organization of the conference.
Supportive community
We are grateful for the opportunity to have organized the 25th ISMIR, and we look forward to continuing to contribute to this thriving and passionate field and its incredibly supportive community of researchers, musicians, and engineers.
To learn more about Adobe Research’s work at the cutting edge of music technology discovery, check out our Audio research page!
