Safety-Constrained Reinforcement Learning with Post-Training Reachability Verification for Robot Navigation

Qisong He, Xinmiao Huang, Jinwei Hu, Zhuoyun Li, Yi Dong, Changshun Wu, Xiaowei Huang.

• Proposed a framework combining CVaR-constrained RL training with neural network reachability verification for safe robot navigation.
• Achieved 98.3% success rate across ten navigation scenarios with the highest safety verification rate.

PrefixGuard: From LLM-Agent Traces to Online Failure-Warning Monitors

Xinmiao Huang, Jinwei Hu, Rajarshi Roy, Changshun Wu, Yi Dong, Xiaowei Huang.

• Preprint on arXiv (May 2026).

Lying with Truths: Open-Channel Multi-Agent Collusion for Belief Manipulation via Generative Montage

Jinwei Hu, Xinmiao Huang, Youcheng Sun, Yi Dong, Xiaowei Huang.

• Accepted to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026).

Chain-of-Thought as a Lens: Evaluating Structured Reasoning Alignment between Human Preferences and Large Language Models

Boxuan Wang, Zhuoyun Li, Xinmiao Huang, Xiaowei Huang, Yi Dong.

• Accepted to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026).

Spatial-DISE: A Unified Benchmark for Evaluating Spatial Reasoning in Vision-Language Models

Xinmiao Huang, Qisong He, Zhenglin Huang, Boxuan Wang, Zhuoyun Li, Guangliang Cheng, Yi Dong, Xiaowei Huang.

• Proposed a unified benchmark, Spatial-DISE, for evaluating spatial reasoning in Vision-Language Models (VLMs).

3D map construction of coal mine roadway mobile robot based on integrated factor graph optimization

ZOU Xiaoyu, HUANG Xinmiao, WANG Zhongbin, FANG Dongsheng, PAN Jie, SI Lei.

• Investigated different popular Lidar-SLAM methods and implemented them on a tracked robot with 3D Lidar and IMU.
• Developed a Lidar-SLAM method based on integrated factor graph optimization.
• Improved the precision and robustness of 3D map construction of coal mine roadways.
• Published one journal paper as the second author.