Jekyll2026-10-10T04:22:56+00:00https://bowen-yan-research.github.io/feed.xmlBowen YanBowen Yan (闫博闻) is a researcher at Shanghai AI Lab working on reasoning and agents, failure-driven self-improvement, and embodied AI. From sensing to reasoning — my research interests2026-10-06T02:00:00+00:002026-10-06T02:00:00+00:00https://bowen-yan-research.github.io/blog/2026/research-overviewMy research connects reasoning and learning with the physical world. I am interested in how embodied agents can understand a task, recognize when execution goes wrong, and use that information to improve their next action.

Reasoning and recovery

My recent work studies VLM-guided recovery from VLA manipulation faults, explicit physical constraint reasoning, and visual communication for robot control. Across these problems, the central question is how to turn an understanding of failure into a targeted correction.

Sensing and embodiment

My earlier work on soft robots and optical sensing involved mechanical design, custom actuators, sensing hardware, finite element analysis, and experiments. This background continues to shape how I think about the relationship between perception, a robot’s body, and its actions.

For project demonstrations, visit my research page. The publication list lists accepted work.

]]>