AsgardBench: A benchmark for visually grounded interactive planning
3月 26, 2026
Imagine a robot tasked with cleaning a kitche…
Microsoft Research ブログ
AsgardBench: A benchmark for visually grounded interactive planning
3月 26, 2026
Imagine a robot tasked with cleaning a kitche…
最近の投稿
-
-
GroundedPlanBench: Spatially grounded long-horizon task planning for robot manipulation
3月 26, 2026Vision-language models (VLMs) use images and …
-
Systematic debugging for AI agents: Introducing the AgentRx framework
As AI agents transition from simple chatbots …
-
PlugMem: Transforming raw agent interactions into reusable knowledge
3月 10, 2026It seems counterintuitive: giving AI agents m…
-
Phi-4-reasoning-vision and the lessons of training a multimodal reasoning model
3月 4, 2026We are pleased to announce Phi-4-reasoning-vi…
-
-
Media Authenticity Methods in Practice: Capabilities, Limitations, and Directions
As synthetic media grows, verifying what’s re…
-
Project Silica’s advances in glass storage technology
2月 18, 2026 | Richard BlackProject Silica introduces new techniques for …
-
Rethinking imitation learning with Predictive Inverse Dynamics Models
2月 5, 2026This research looks at why Predictive Inverse…
-
Paza: Introducing automatic speech recognition benchmarks and models for low resource languages
Microsoft Research unveils Paza, a human-cent…
-
UniRG: Scaling medical imaging report generation with multimodal reinforcement learning
1月 27, 2026AI can help generate medical image reports, b…
-
Argos: Multimodal reinforcement learning with agentic verifier for AI agents
1月 20, 2026Argos improves multimodal RL by evaluating wh…
Explore More
-
Events & conferences
Meet our community of researchers, learn about exciting research topics, and grow your network
-
Podcasts
Ongoing conversations at the cutting edge of research
-
Microsoft Research Forum
Join us for a continuous exchange of ideas about research in the era of general AI