AI Radar Research

Daily research digest for developers — Saturday, March 28 2026

Microsoft Research AI

GroundedPlanBench: Spatially grounded long-horizon task planning for robot manipulation

Vision-language models (VLMs) use images and text to plan robot actions, but they still struggle to decide what actions to take and where to take them. This research introduces a benchmark for evaluating spatially grounded long-horizon task planning in robotics.

Why it matters: This benchmark provides a framework to assess and improve the planning capabilities of AI systems in complex, real-world environments.
OpenAI Blog

Creating with Sora Safely

OpenAI has developed Sora 2 and the Sora app to address safety challenges in video models and social creation platforms. The approach focuses on building safety into the foundation of these systems.

Why it matters: Ensuring safety in AI systems is crucial for their reliable deployment in creative and social contexts.
✉ Subscribe to daily research digest