From raw interaction to reusable knowledge: Rethinking memory for AI agents
It seems counterintuitive: giving AI agents more memory can make them less effective. As interaction logs accumulate, they grow large, fill with irrelevant content, and become increasingly difficult to use. More memory means that agents must search through larger volumes of past interactions to find...
Phi-4-reasoning-vision and the lessons of training a multimodal reasoning model
We are pleased to announce Phi-4-reasoning-vision-15B, a 15 billion parameter open‑weight multimodal reasoning model, available through Microsoft Foundry (opens in new tab), HuggingFace (opens in new tab) and GitHub (opens in new tab). Phi-4-reasoning-vision-15B is a broadly capable model that can b...
By mid-morning, a typical knowledge worker is already juggling a client report, a budget spreadsheet, a slide deck, and an email backlog, all interdependent and all demanding attention at once. For AI agents to be genuinely useful in that environment, they will need to operate the same way, but toda...
Faster decisions: How an AI agent is redefining executive workflows at one of the world’s largest building materials companies
The post Faster decisions: How an AI agent is redefining executive workflows at one of the world’s largest building materials companies appeared first on Source.
Rethinking imitation learning with Predictive Inverse Dynamics Models
This research looks at why Predictive Inverse Dynamics Models often outperform standard Behavior Cloning in imitation learning. By using simple predictions of what happens next, PIDMs reduce ambiguity and learn from far fewer demonstrations.
The post Rethinking imitation learning with Predictive Inv...
Paza: Introducing automatic speech recognition benchmarks and models for low resource languages
Microsoft Research unveils Paza, a human-centered speech pipeline, and PazaBench, the first leaderboard for low-resource languages. It covers 39 African languages and 52 models and is tested with communities in real settings.
The post Paza: Introducing automatic speech recognition benchmarks and mo...
Multimodal reinforcement learning with agentic verifier for AI agents
Argos improves multimodal RL by evaluating whether an agent’s reasoning aligns with what it observes over time. The approach reduces visual hallucinations and produces more reliable, data-efficient agents for real-world applications.
The post Multimodal reinforcement learning with agentic verifier f...
OptiMind: A small language model with optimization expertise
OptiMind is a small language model that converts business operation challenges, described naturally, into mathematical formulations that optimization software can solve. It reduces formulation time & errors & enables fast, privacy-preserving local use.
The post OptiMind: A small language model with ...
Agent Lightning: Adding reinforcement learning to AI agents without code rewrites
By decoupling how agents work from how they’re trained, Agent Lightning turns each step an agent takes into data for reinforcement learning. This makes it easy for developers to improve agent performance with almost zero code changes.
The post Agent Lightning: Adding reinforcement learning to AI age...
Promptions helps make AI prompting more precise with dynamic UI controls
Promptions helps developers add dynamic, context-aware controls to chat interfaces so users can guide generative AI responses. It lets users shape outputs quickly without writing long instructions.
The post Promptions helps make AI prompting more precise with dynamic UI controls appeared first on Mi...
GigaTIME: Scaling tumor microenvironment modeling using virtual population generated by multimodal AI
Using AI-generated virtual populations, Microsoft researchers uncovered hidden cellular patterns that could reshape how we understand and treat cancer.
The post GigaTIME: Scaling tumor microenvironment modeling using virtual population generated by multimodal AI appeared first on Microsoft Research...
Ideas: Community building, machine learning, and the future of AI
As the Women in Machine Learning Workshop (WiML) marks its 20th annual gathering, cofounders, friends, and collaborators Jenn Wortman Vaughan and Hanna Wallach reflect on WiML’s evolution, navigating the field of ML, and their work in responsible AI.
The post Ideas: Community building, machine learn...
Reducing Privacy leaks in AI: Two approaches to contextual integrity
New research explores two ways to give AI agents stronger privacy safeguards grounded in contextual integrity. One adds lightweight, inference-time checks; the other builds contextual awareness directly into models through reasoning and RL.
The post Reducing Privacy leaks in AI: Two approaches to co...