5 papers
RegenHarness: A Robot Agent Harness with Evidence-Gated Recursive Self-Improvement
Kailin Wang, Haoxiang Jie, Yaoyuan Yan +2
Long-horizon robot execution requires a clear distinction between a model's proposal, a controller's termination, and verified task completion. We present RegenHarness, an evidence…
Harness Robotic OS: A Unified Embodied-Agent Runtime for Closed-Loop Quadruped Inspection
Yaoyuan Yan, Zhiyou Heng, Haoxiang Jie +3
Autonomous property inspection requires more than robust robot navigation: a deployable system must connect heterogeneous sensing, reusable autonomy capabilities, multimodal scene…
Ontology-Grounded World Models for Failure Diagnosis and Closed-Loop Repair in Physical AI Systems
Kailin Wang, Haoxiang Jie, Yaoyuan Yan +2
EV-WM represents candidate quality with feature and event scores, but these scores do not explicitly record an unmet task predicate, a route label for an available correction mecha…
EV-WM: Event-Verified World Models for Long-Horizon Robotic Manipulation
Kailin Wang, Haoxiang Jie, Yaoyuan Yan +2
Pretrained-feature world models provide a useful substrate for robot imagination, but visual or latent prediction alone does not determine whether an imagined future satisfies task…
OmniVLA-RL: A Vision-Language-Action Model with Spatial Understanding and Online RL
Haoxiang Jie, Yaoyuan Yan, Xiangyu Wei +4
Visual-Language-Action (VLA) models represent a paradigm shift in embodied AI, yet existing frameworks often struggle with imprecise spatial perception, suboptimal multimodal fusio…