Mi Yan 严汨

Scene & object generalization

GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Data

Shengliang Deng*, Mi Yan*, Songlin Wei, Haixin Ma, Yuxin Yang, Jiayi Chen, Zhiqi Zhang, Taoyu Yang, Xuheng Zhang, Wenhao Zhang, Heming Cui, Zhizheng Zhang, He Wang

Our first simulation data-generation and evaluation pipeline produces an open billion-frame dataset of interactions with 10,000 objects, enabling zero-shot sim-to-real grasping without real-world fine-tuning.

Camera generalization

StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision

Shengliang Deng*, Mi Yan*, Yixin Zheng*, Jiayi Su, Wenhao Zhang, Xiaoguang Zhao, Heming Cui, Zhizheng Zhang, He Wang

Building on our simulation pipeline, StereoVLA extracts rich geometry from large-scale synthetic stereo-action data through GeoSem and 2 synergistic co-training tasks, enabling robust manipulation across near-hemispherical camera viewpoints.

Task generalization

GIF: Agentic Generation of Interactive and Functional Object Compositions for Robot Learning

Long Xu*, Zhiqi Zhang*, Mi Yan*, S. Deng, C. Xia, M. Dong, J. Chen, J. Lyu, F. Gao, Z. Zhang, H. Wang

From a task instruction, GIF automatically generates interactive, functional scenes and robot trajectories, synthesizing data to train and evaluate VLA policies that transfer to real robots.

Embodiment generalization

ZETA: A Controlled Study of Zero-Shot Cross-Embodiment VLA Transfer for Tabletop Manipulation

Mi Yan*, Wenhao Zhang*, Zhiqi Zhang*, Yu Peng*, Tangxinyu Wang*, Lingfei Zhai, Jiayi Su, Shengliang Deng, Lin Peng, Yaowei Liu, Yuxing Chen, Zhiyuan Wei, Jilong Wang, Jiayi Chen, Jiangran Lyu, Zhizheng Zhang, He Wang

Using a simulated dataset spanning 512 embodiments, ZETA systematically analyzes cross-embodiment VLA transfer across state-action representations, embodiment diversity, co-training tasks, and zero-shot definitions, distinguishing strict transfer from transfer with target-embodiment pretraining exposure.

LDA-1B: Scaling Latent Dynamics Action Model via Universal Embodied Data Ingestion

Jiangran Lyu*, Kai Liu*, Xuheng Zhang*, Haoran Liao, Yusen Feng, Wenxuan Zhu, Tingrui Shen, Jiayi Chen, Jiazhao Zhang, Yifei Dong, Wenbo Cui, Senmao Qi, Shuo Wang, Yixin Zheng, Mi Yan, Xuesong Shi, Haoran Li, Dongbin Zhao, Ming-Yu Liu, Zhizheng Zhang, Li Yi, Yizhou Wang, He Wang

Unifies dynamics, visual prediction, and policy learning from 30,000+ hours of heterogeneous human and robot interaction data.

Any3D-VLA: Enhancing VLA Robustness via Diverse Point Clouds

Xianzhe Fan, Shengliang Deng, Xiaoyang Wu, Yuxiang Lu, Zhuoling Li, Mi Yan, Yujia Zhang, Zhizheng Zhang, He Wang

Combines simulated, sensor-captured, and model-estimated point clouds to make VLA policies robust to depth-scale and environmental shifts.

Emerging Extrinsic Dexterity in Cluttered Scenes via Dynamics-aware Policy Learning

Yixin Zheng*, Jiangran Lyu*, Yifan Zhang, Jiayi Chen, Mi Yan, Yuntian Deng, Xuesong Shi, Xiaoguang Zhao, Yizhou Wang, Zhizheng Zhang, He Wang

Learns contact dynamics to guide reinforcement learning, enabling robots to exploit environmental contact for manipulation in cluttered scenes.

KPGrasp: Scalable Keypoint Flow Matching for Dexterous Grasp Generation

Yuansen Huang*, Jiayi Chen*, Haoran Liu, Yubin Ke, Bing Han, Jiangran Lyu, Mi Yan, Li Yi, He Wang

Represents dexterous grasps as 3D hand keypoints and learns a scalable flow-matching prior without contact losses or contact-based test-time refinement.

Adaptive Zone-aware Hierarchical Planner for Vision-Language Navigation

Chen Gao, Xingyu Peng, Mi Yan, He Wang, Lirong Yang, Haibing Ren, Hongsheng Li, Si Liu

Decomposes instruction-following navigation into adaptive zone-level subgoal planning and low-level execution for long-horizon navigation.

MaskClustering: View Consensus based Mask Graph Clustering for Open-Vocabulary 3D Instance Segmentation

Mi Yan, Jiazhao Zhang, Yan Zhu, He Wang

Clusters 2D masks using multi-view consensus to recover open-vocabulary 3D instances without training.

Tracking and Reconstructing Hand Object Interactions from Point Cloud Sequences in the Wild

Jiayi Chen*, Mi Yan*, Jiazhao Zhang, Yinzhen Xu, Xiaolong Li, Yijia Weng, Li Yi, Shuran Song, He Wang

Tracks hand–object poses and reconstructs object shapes from real-world point-cloud sequences using models trained on synthetic data.

Domain Adaptation on Point Clouds via Geometry-Aware Implicits

Yuefan Shen, Yanchao Yang, Mi Yan, He Wang, YouYi Zheng, Leonidas Guibas

Uses geometry-aware implicit reconstruction as self-supervision to bridge domain gaps in point-cloud recognition.