5 papers indexed
Xiao Yu, Baolin Peng, Ruize Xu, Hao Zou, Qianhui Wu, Hao Cheng, et al.
Modern AI agents rely on elaborate inference harnesses such as Claude Code, Codex, and OpenClaw to drive multi-turn reasoning, tool use, and access to external systems. While powerful, these complex harnesses also make agents hard to train end-to-end with open infrastructure, who…
Feng Wang, Hao Cheng, Y Cao, Guanyu Hu, F Wang, Xiao Ma, et al.
Abstract Track irregularity reflects deviations in track geometry and serves as a key indicator of rail-way infrastructure performance. Accurate evaluation and prediction of track quality are cru-cial for detecting potential safety risks and guiding maintenance decisions. Traditi…
Leitian Tao, Baolin Peng, Wenlin Yao, Tao Ge, Hao Cheng, Mike Hang Wang, et al.
Multi-turn agents solve complex tasks through extended sequences of tool interactions before producing a final answer, making credit assignment a fundamental challenge during post-training. Outcome rewards provide reliable supervision for short-horizon reasoning, but become spars…
Zishuo Li, Bowen Yang, Changtao Miao, Kai Zhu, Hao Chen, Qingze Guan, et al.
Egocentric videos of human manipulation provide scalable supervision for embodied intelligence, yet existing resources rarely combine low-cost continuous capture, manipulation-level structured annotations, and reusable tools for robot learning. We present Open-AoE, an open, commu…
Mengmeng Liu, Diankun Zhang, Jiuming Liu, Jianfeng Cui, Hongwei Xie, Guang Chen, et al.
World Action Models (WAMs) have shown strong potential for improving action generalization in autonomous driving by using future video prediction as dense supervision for scene dynamics and temporal causality. However, it remains unclear which architecture better transfers video-…