Xianhui Meng

M.S. Candidate

University of Science and Technology of China (USTC)

Biography

I am currently a M.S. student in University of Science and Technology of China (USTC), advised by Prof. Jun Liu. My research interests focus on computer vision, Autonomous Driving, Embodied AI and Agent.

I am a strong advocate of sharing🌱, collaborating🤝, advancing🚀, and innovating💡. Passionate about bridging fundamental research with real-world impact, I strive to ensure my work contributes to both academic progress and practical value.

Collaboration welcome! If you are interested in collaborating or would like to connect, feel free to send me an email.

Current Interests

Computer Vision
Autonomous Driving
Embodied AI
Agent

Services

Program Committee Member: ICME 2025, CVPR 2026, etc.

Education

M.S. in Information and Communication Engineering, 2024.09 – Present
University of Science and Technology of China
B.S. in Automation, 2020.09 – 2024.07
Southeast University

Services

Program Committee Member: ICME 2025, CVPR 2026, etc.

Publications & Preprints

Full publication and BibTeX details are also available on Google Scholar.

MiMo-Embodied: X-Embodied Foundation Model Technical Report

Technical Report

The first cross-embodied foundation model to integrate autonomous driving and embodied AI, setting new records across 17 embodied and 12 driving benchmarks.

Xiaoshuai Hao, Lei Zhou, Zhijian Huang, Zhiwen Hou, Yingbo Tang, Lingfeng Zhang, Guang Li, Zheng Lu, Shuhuai Ren, Xianhui Meng, Yuchen Zhang, et al.

Paper Project

PRPO: Aligning Process Reward with Outcome Reward in Policy Optimization

arXiv 2026

Combines outcome rewards with process-level guidance for critic-free policy optimization, improving multi-step LLM reasoning and credit assignment.

Ruiyi Ding, Yongxuan Lv, Xianhui Meng, Jiahe Song, Chao Wang, Chen Jiang, Yuan Cheng

Paper

DSBench: A Comprehensive Benchmark for Evaluating External and In-Cabin Risks

ACM MM 2026

A unified driving-safety benchmark covering 10 external-risk and in-cabin categories, with 98K instances for evaluating and improving VLM safety.

Xianhui Meng, Yuchen Zhang, Zhijian Huang, Zheng Lu, Ziling Ji, Yandan Lin, Yaoyao Yin, Hongyuan Zhang, Wei Zhou, et al.

Paper Project

The Evaluation Bottleneck of Vision-Language-Action Models: A Evaluation-Centric Survey

Preprint 2026

A survey of VLA evaluation that identifies key benchmarking bottlenecks and systematically examines how existing protocols measure vision-language-action models.

Zirui Song, Huaxing Liu, Xiang Wang, Shuai Li, Xinye Li, Yuheng Ji, Lang Gao, Jinghui Zhang, Xianhui Meng, Xiaojun Chang, et al.

Paper Project

The Evaluation Bottleneck of Vision-Language-Action Models

Probing Effective and Efficient Category-Level Articulated Object Pose Perception

TPAMI 2026

Studies category-level articulated-object pose perception and introduces a rotation-decoupled strategy for accurate, efficient 6D pose estimation and tracking.

Li Zhang, Xianhui Meng, Liu Liu, Haonan Jiang, Jianan Wang, Rujing Wang, Cewu Lu, Jun Liu, Hong Zhang

Paper Project

Probing Effective and Efficient Category-Level Articulated Object Pose Perception

DICArt: Advancing Category-level Articulated Object Pose Estimation in Discrete State-Spaces

CVPR 2026

Models category-level articulated-object pose estimation in discrete state spaces, improving pose accuracy and robustness for articulated objects.

Li Zhang, Mingyu Mei, Ailing Wang, Xianhui Meng, Yan Zhong, Xinyuan Song, Liu Liu, Rujing Wang, Zaixing He, Cewu Lu

Paper Project

Diff-Art: Category-level Articulation Pose Estimation via Conditional Diffusion

ICME 2025

Estimates category-level articulation pose with a conditional diffusion model, capturing uncertainty and complex articulation distributions for 6D pose estimation.

Yukang Huo, Xianhui Meng, Li Zhang†, Haonan Jiang, Yan Zhong, Mingyuan Yao, Haihua Wang†

Paper

Text-Driven 3D Indoor Scene Synthesis in Non-Manhattan Environments

arXiv 2026

Generates physically plausible, text-driven indoor scenes in non-Manhattan rooms using statistical layout priors and hierarchical object placement.

Xianhui Meng, Zirui Song, Yuchen Zhang, Li Zhang, Yongxuan Lv, Xiuying Chen, Kun Wang, Yan Luo, Kai Chen, Hangjun Ye, et al.

Paper

Text-Driven 3D Indoor Scene Synthesis in Non-Manhattan Environments

Robust 3D Alignment of Generative Reconstructions via Partial Monocular Observations

arXiv 2026

A training-free Sim(3) alignment framework with coarse-to-fine matching and hallucination filtering for grounding generative 3D priors in partial monocular observations.

Yuchen Zhang, Luanyuan Dai, Yiwei Wang, Xiwei Xu, Jianing Zhang, Xianhui Meng, Yanbiao Ma, Jiayi Ma, Xiaoshuai Hao

Paper

Robust 3D Alignment of Generative Reconstructions via Partial Monocular Observations

Exploring Category-level Articulated Object Pose Tracking on SE(3) Manifolds

AAAI 2026

Formulates articulated-object tracking on SE(3) manifolds and uses SE(3)-invariant pose-voting parameters for robust category-level pose tracking.

Xianhui Meng, Yukang Huo, Li Zhang, Liu Liu, Haonan Jiang, Yan Zhong, Pingrui Zhang, Cewu Lu, Jun Liu

Paper Project

Exploring Category-level Articulated Object Pose Tracking on SE(3) Manifolds

Xianhui Meng

M.S. Candidate

University of Science and Technology of China (USTC)

Biography

Current Interests

Services

Education

Recent News

Research Interests

Services

Publications & Preprints

MiMo-Embodied: X-Embodied Foundation Model Technical Report

PRPO: Aligning Process Reward with Outcome Reward in Policy Optimization

DSBench: A Comprehensive Benchmark for Evaluating External and In-Cabin Risks

The Evaluation Bottleneck of Vision-Language-Action Models: A Evaluation-Centric Survey

Probing Effective and Efficient Category-Level Articulated Object Pose Perception

DICArt: Advancing Category-level Articulated Object Pose Estimation in Discrete State-Spaces

Diff-Art: Category-level Articulation Pose Estimation via Conditional Diffusion

Text-Driven 3D Indoor Scene Synthesis in Non-Manhattan Environments

Robust 3D Alignment of Generative Reconstructions via Partial Monocular Observations

Exploring Category-level Articulated Object Pose Tracking on SE(3) Manifolds

Contact