Rapid developments in large language models and generative audio systems raise questions about the relationship between industrial innovation and academic research in music information retrieval (MIR), particularly when it comes with large foundation models. How closely do the questions, methods, and evaluation practices represented in academic publications reflect those driving industrial development? As training these large models at scale demands substantial resources, how can university labs contribute to the industrial landscape, and how should they prepare students?

In this majlis, we will explore where industry and academia diverge, where they already collaborate, and what each can learn from the other. Alongside the challenges of access to compute, data, and models, we will discuss the distinctive contributions academic labs can make through fundamental research, musical and cultural expertise, independent evaluation, efficient methods, and open research resources. We will then turn to a practical question: how can our community build stronger connections through shared research agendas, access to industrial systems, collaborative benchmarks, and training opportunities that benefit both sectors?

Panelists

Xubo Liu Meta Superintelligence Labs Website	Stefan Lattner Sony CSL (Paris) Website	Shuqi Dai xAI Website

Moderator

Ajay Srinivasamurthy Amazon Alexa (Bengaluru) Website