RecSys 2026 Challenge

The RecSys 2026 Challenge on Conversational Music Recommendation (Music-CRS) is coordinated by Seungheon Doh (KAIST, South Korea), Sergio Oramas (Pandora/SiriusXM), Bruno Sguerra (Deezer Research), Abhinav Bohra (Amazon), Claudio Pomo (Politecnico di Bari, Italy), and Francesco Barile (Maastricht University, the Netherlands).

The challenge focuses on the development of conversational recommender systems that can understand user music preferences through natural, multi-turn dialogue, recommend relevant tracks from a music catalog, and generate coherent and informative natural language responses. In this setting, participants are asked to build systems that take the dialogue history and user context as input and produce both ranked track recommendations and response text. The task is based on the TalkPlayData dataset, which includes multi-turn conversations, track metadata, and user profiles, enabling participants to focus on modeling and evaluation.

The evaluation is designed to reflect both recommendation quality and response quality. The shared task materials identify ranking quality as a core evaluation component, while the overall challenge framing emphasizes the importance of interactive, personalized, and explainable recommendation behavior. A detailed description of the task, dataset resources, baseline materials, and submission format is available on the challenge website. The RecSys 2026 Challenge will take place in conjunction with the 20th ACM Conference on Recommender Systems, to be held in Minneapolis, Minnesota, USA, from September 28 to October 2, 2026.