We present Phi-4-reasoning-vision-15B, a compact open-weight multimodal reasoning model, and share the motivations, design choices, experiments, and learnings that informed its development. Our goal is to contribute practical insight to the research community on building smaller, efficient multimodal reasoning models and to share the result…
MSR Synth is a research project exploring how to design AI agents with social intelligence—agents that can communicate, collaborate, and participate in human interaction in ways that are context‑aware, trustworthy, and aligned with human values. Drawing on human–computer interaction, AI‑mediated communication, and social computing, the…
By Ben Finkelshtein (opens in new tab) (University of Oxford), Silviu Cucerzan, Sujay Kumar Jauhar, and Ryen W. White (Microsoft) Think about the last time you opened a shared document at work. Behind that simple action lies a complex network of relationships: the colleagues who edited the file before you, the team site on…
Brain-Computer Interfaces (BCIs) are moving beyond their original and main applications as rehabilitation technologies, to become promising tools for interacting with users by measuring their mental states. This talk presents an overview of modern BCIs and focuses on their integration with immersive technologies such as…
Dion2 reduces the cost of Muon’s orthonormalization step by orthonormalizing only a small, selected submatrix at each iteration. This lightweight approach preserves Muon’s strong performance while significantly improving scalability of optimizer at scale.
We present Adaptively Rotated Optimization (ARO), a matrix optimizer that speeds up LLM training by applying updates in a rotated, geometry-aware coordinate system. Guided by new insights on global structures on LLM loss landscapes, ARO treats rotation as a unifying principle for sample efficiency, and…
The Tyger framework enables faster, more accessible medical imaging by streaming raw data to the cloud for accelerated reconstruction—reducing patient wait times and discomfort—while empowering researchers to rapidly test and deploy new algorithms.
OptiMind is a specialized language model that translates natural-language problem descriptions directly into solver-ready mathematical optimization formulations. This removes one of the most expertise-intensive bottlenecks in optimization workflows and makes advanced optimization more accessible.
Agent Lightning is an agent optimization framework that enables agents to learn from their experiences through reinforcement learning and other methods. By treating agents as first-class citizens, optimization becomes automatic for any agent with minimal code changes.
As AI agents move from isolated tools to active participants in multi-agent ecosystems, their success depends on more than task competence—it requires strategic behavior under misaligned incentives and imperfect information. Using Magentic Marketplace, an open-source simulation of two-sided agent markets, we show that while frontier…
Technical advances are moving at such a rapid pace that it can be challenging to define the tomorrow we’re working toward. In The Shape of Things to Come, Microsoft Research leader Doug Burger and experts from across disciplines tease out the thorniest AI issues facing…
Microsoft research lead Doug Burger introduces his new podcast series, "The Shape of Things to Come", an exploration into the fundamental truths about AI and how the technology will reshape the future.
