CORTEXA
← Browse

Weixin Li

2 papers indexed

arxivcs.CVcs.AIcs.CLcs.GRcs.MA2026-07-17

Clarify Before Executing: A Self-Evolving Agent for Resolving Intent Asymmetry in 3D Tool Orchestration

Xiaoye Zhu, Weixin Li, Junan Huo, Bozhong Wang, Jia Zeng, Yi Yang, et al.

A fundamental intent asymmetry plagues modern 3D asset creation: while state-of-the-art 3D toolchains demand precise, executable parameters, ordinary users typically provide vague, underspecified instructions. Current 3D agents treat this ambiguity as noise, defaulting to blind e…

View free PDFSource page
arxivcs.LGcs.AIcs.RO2026-07-14

TerraZero: Procedural Driving Simulation for Zero-Demonstration Self-Play at Scale

Zhouchonghao Wu, Akshay Rangesh, Weixin Li, Wei-Jer Chang, Zachary Lee, Tim Wang, et al.

Training robust autonomous driving agents requires a simulator that is fast enough for reinforcement learning at scale, realistic enough to ground behavior in real-world map structure, and diverse enough to cover the safety-critical long tail that logged data rarely contains. We…

View free PDFSource page