Sihang LI

2026

1. Arxiv

   FLoRe: Learning Force-Reactive Humanoid Loco-Manipulation

   Sihang Li, Xiaofeng Guo, Jinkun Cao, and 5 more authors

   2026

   Abs arXiv PDF Code Website

   Many everyday tasks require controlling contact force as well as motion. On a humanoid, this force must often be coordinated across both hands and the whole body, and cannot be specified by kinematic trajectories alone. Learning such tasks by imitation therefore needs demonstrations that record contact force, and a whole-body controller that can deliver it. We present FLoRe, a framework for learning force-reactive humanoid loco-manipulation. Its robot-free force-annotated data collection system records the contact wrench applied through two UMI-style handheld grippers, together with hand, torso, and foot motion. A force-reactive diffusion policy takes vision, proprioception, and measured wrenches as input, then predicts actions and per-hand force commands. A hybrid compliance whole-body controller with force tracking executes these commands while maintaining balance: an integral of the force error measured by real force/torque (F/T) sensors lets each hand track force accurately along the commanded direction and comply along the others. In simulation, the controller tracks per-hand force commands of varied magnitude and direction, with little loss in whole-body motion accuracy. Across three contact-rich tasks on the robot, the force-reactive policy succeeds in 80% of rollouts, compared with 26.7% for a position-only policy trained on the same demonstrations and executed by the same controller.

2. Arxiv

   RoboDoctor: Training-Free Recursive Self-Improvement for Robot Policies

   Heyang Yu*, Haomin Li*, Sihang Li, and 7 more authors

   2026

   Abs arXiv PDF Code Website

   Generalist robot policies and language-model agents provide complementary capabilities, but completing a task also requires adapting their use to the consequences of physical interaction. We study execution as a locus of adaptation: feedback guides the current action and supplies evidence for revising subsequent decisions under fixed model weights. We introduce RoboDoctor, a training-free framework that connects diagnosis, grounded action, and persistent experience across three revision targets. The harness allocates control between policy and agent; intervention determines when to assist while preserving useful policy behavior; and operation refines physical methods through observed effects. Linked experience records retain the reason for a decision, the attempted method, its outcome, and conditions for reuse. On a curated archive of 54 RoboDojo-Sim task entries, RoboDoctor achieves a five-dimension mean score of 57.10, with a native success rate of 49.44%. Recorded interactions illustrate strategy changes, selective correction, and grasp revision. These findings characterize execution-time adaptation in the evaluated system; controlled trials remain necessary to isolate the contribution of cross-episode experience.

3. ECCV

   

   SOMO: Scalable Cross-embodiment Dexterous Grasping via Morphology-Prior Diffusion

   Sihang Li, Zheming Zhou, Omid Ghasemalizadeh, and 4 more authors

   European Conference on Computer Vision, 2026

   Abs PDF

   This paper presents SOMO, a scalable framework for cross-embodiment grasp synthesis that transfers to novel robot hands using only their hand description (i.e., a Unified Robot Description Format (URDF) file), without requiring any hand–object interaction annotations. Unlike prior approaches that rely on hand-specific models or annotated grasp data for each embodiment, SOMO introduces a shared Morphology-Prior Diffusion model applicable across heterogeneous hands through three key designs. First, grasping is formulated as a 3D assembly problem by predicting per-link SE(3) poses, enabling geometry-driven generation independent of hand-specific kinematics, with feasibility enforced through a post joint optimization stage. Second, a classifier-free training strategy learns a morphology prior from physically valid hand configurations generated through forward-kinematic exploration without object–grasp annotations, enabling grasp synthesis for unseen robot hands given only their URDF files. Third, a 3D shape-aware VAE encodes link and object geometry into a shared embedding, enabling consistent reasoning about hand–object complementarity across embodiments. Experiments show that SOMO achieves state-of-the-art grasp synthesis across six robot hands and demonstrates strong annotation-free generalization to previously unseen hands.

4. Arxiv

   

   EgoPush: Learning End-to-End Egocentric Multi-Object Rearrangement for Mobile Robots

   Boyuan An, Zhexiong Wang, Yipeng Wang, and 4 more authors

   2026

   Abs arXiv PDF Code Website

   Humans can rearrange objects in cluttered environments using egocentric perception, navigating occlusions without global coordinates. Inspired by this capability, we study long-horizon multi-object non-prehensile rearrangement for mobile robots using a single egocentric camera. We introduce EgoPush, a policy learning framework that enables egocentric, perception-driven rearrangement without relying on explicit global state estimation that often fails in dynamic scenes. EgoPush designs an object-centric latent space to encode relative spatial relations among objects, rather than absolute poses. This design enables a privileged reinforcement-learning (RL) teacher to jointly learn latent states and mobile actions from sparse keypoints, which is then distilled into a purely visual student policy. To reduce the supervision gap between the omniscient teacher and the partially observed student, we restrict the teacher’s observations to visually accessible cues. This induces active perception behaviors that are recoverable from the student’s viewpoint. To address long-horizon credit assignment, we decompose rearrangement into stage-level subproblems using temporally decayed, stage-local completion rewards. Extensive simulation experiments demonstrate that EgoPush significantly outperforms end-to-end RL baselines in success rate, with ablation studies validating each design choice. We further demonstrate zero-shot sim-to-real transfer on a mobile platform in the real world.

5. ICML

   

   CRAG: Can 3D Generative Models Help 3D Assembly?

   Zeyu Jiang*, Sihang Li*, Siqi Tan, and 8 more authors

   2026

   Abs arXiv PDF Code Website

   Most existing 3D assembly methods treat the problem as pure pose estimation, rearranging observed parts via rigid transformations. In contrast, human assembly naturally couples structural reasoning with holistic shape inference. Inspired by this intuition, we reformulate 3D assembly as a joint problem of assembly and generation. We show that these two processes are mutually reinforcing: assembly provides part-level structural priors for generation, while generation injects holistic shape context that resolves ambiguities in assembly. Unlike prior methods that cannot synthesize missing geometry, we propose CRAG, which simultaneously generates plausible complete shapes and predicts poses for input parts. Extensive experiments demonstrate state-of-the-art performance across in-the-wild objects with diverse geometries, varying part counts, and missing pieces. Our code and models will be released.

2025

1. ICCV

   

   GARF: Learning Generalizable 3D Reassembly for Real-World Fractures

   Sihang Li, Jiang Zeyu, Chen Grace, and 9 more authors

   International Conference on Computer Vision, 2025

   Abs arXiv PDF Code Website

   3D reassembly is a challenging spatial intelligence task with broad applications across scientific domains. While large-scale synthetic datasets have fueled promising learning-based approaches, their generalizability to different domains is limited. Critically, it remains uncertain whether models trained on synthetic datasets can generalize to real-world fractures where breakage patterns are more complex. To bridge this gap, we propose GARF, a generalizable 3D reassembly framework for real-world fractures. GARF leverages fracture-aware pretraining to learn fracture features from individual fragments, with flow matching enabling precise 6-DoF alignments. At inference time, we introduce one-step preassembly, improving robustness to unseen objects and varying numbers of fractures. In collaboration with archaeologists, paleoanthropologists, and ornithologists, we curate Fractura, a diverse dataset for vision and learning communities, featuring real-world fracture types across ceramics, bones, eggshells, and lithics. Comprehensive experiments have shown our approach consistently outperforms state-of-the-art methods on both synthetic and real-world datasets, achieving 82.87% lower rotation error and 25.15% higher part accuracy. This sheds light on training on synthetic data to advance real-world 3D puzzle solving, demonstrating its strong generalization across unseen object shapes and diverse fracture types.

2. ICCV

   

   Adversarial Exploitation of Data Diversity Improves Visual Localization

   Sihang Li, Siqi Tan, Bowen Chang, and 3 more authors

   International Conference on Computer Vision, 2025

   Abs arXiv PDF Code Website

   Visual localization, which estimates a camera’s pose within a known scene, is a long-standing challenge in vision and robotics. Recent end-to-end methods that directly regress camera poses from query images have gained attention for fast inference. However, existing methods often struggle to generalize to unseen views. In this work, we aim to unleash the power of data synthesis to promote the generalizability of pose regression. Specifically, we lift real 2D images into 3D Gaussian Splats with varying appearance and deblurring abilities, which are then used as a data engine to synthesize more posed images. To fully leverage the synthetic data, we build a two-branch joint training pipeline, with an adversarial discriminator to bridge the syn-to-real gap. Experiments on established benchmarks show that our method outperforms state-of-the-art end-to-end approaches, reducing translation and rotation errors by 50% and 21.6% on indoor datasets, and 35.56% and 38.7% on outdoor datasets. We also validate the effectiveness of our method in dynamic driving scenarios under varying weather conditions. Notably, as data synthesis scales up, our method exhibits a growing ability to interpolate and extrapolate training data for localizing unseen views.

2024

1. NeurIPS

   

   Multiview Scene Graph

   Juexiao Zhang, Gao Zhu, Sihang Li, and 4 more authors

   In Advances in Neural Information Processing Systems, 2024

   Abs arXiv PDF Code Website

   A proper scene representation is central to the pursuit of spatial intelligence where agents can robustly reconstruct and efficiently understand 3D scenes. A scene representation is either metric, such as landmark maps in 3D reconstruction, 3D bounding boxes in object detection, or voxel grids in occupancy prediction, or topological, such as pose graphs with loop closures in SLAM or visibility graphs in SfM. In this work, we propose to build \textitMultiview Scene Graphs (MSG) from unposed images, representing a scene topologically with interconnected place and object nodes. The task of building MSG is challenging for existing representation learning methods since it needs to jointly address both visual place recognition, object detection, and object association from images with limited fields of view and potentially large viewpoint changes. To evaluate any method tackling this task, we developed an MSG dataset and annotation based on a public 3D dataset. We also propose an evaluation metric based on the intersection-over-union score of MSG edges. Moreover, we develop a novel baseline method built on mainstream pretrained vision models, combining visual place recognition and object association into one Transformer decoder architecture. Experiments demonstrate our method has superior performance compared to existing relevant baselines.

2. IROS

   

   SSCBench: A Large-Scale 3D Semantic Scene Completion Benchmark for Autonomous Driving

   Yiming Li, Sihang Li, Xinhao Liu, and 8 more authors

   In IEEE/RSJ International Conference on Intelligent Robots and Systems, 2024

   Abs arXiv PDF Code

   Semantic scene completion (SSC) is crucial for holistic 3D scene understanding by jointly estimating semantics and geometry from sparse observations. However, progress in SSC, particularly in autonomous driving scenarios, is hindered by the scarcity of high-quality datasets. To overcome this challenge, we introduce SSCBench, a comprehensive benchmark that integrates scenes from widely-used automotive datasets (e.g., KITTI-360, nuScenes, and Waymo). SSCBench follows an established setup and format in the community, facilitating the easy exploration of the camera- and LiDAR-based SSC across various real-world scenarios. We present quantitative and qualitative evaluations of state-of-the-art algorithms on SSCBench and commit to continuously incorporating novel automotive datasets and SSC algorithms to drive further advancements in this field.

2023

1. RA-L

   

   Boosting UAV Tracking with Voxel-based Trajectory-Aware Pre-Training

   Sihang Li, Changhong Fu, Kunhan Lu, and 3 more authors

   IEEE Robotics and Automation Letters, 2023

   Abs PDF Code

   Siamese network-based object tracking has remarkably promoted the automatic capability for highlymaneuvered unmanned aerial vehicles (UAVs). However, the leading-edge tracking framework often depends on template matching, making it trapped when facing multiple views of object in consecutive frames. Moreover, the general imagelevel pretrained backbone can overfit to holistic representations, causing the misalignment to learn object-level properties in UAV tracking. To tackle these issues, this work presents TRTrack, a comprehensive framework to fully exploit the stereoscopic representation for UAV tracking. Specifically, a novel pretraining paradigm method is proposed. Through trajectoryaware reconstruction training, the capability of the backbone to extract stereoscopic structure feature is strengthened without any parameter increment. Accordingly, an innovative hierarchical self-attention Transformer is proposed to capture the local detail information and global structure knowledge. For optimizing the correlation map, we proposed a novel spatial correlation refinement (SCR) module, which promotes the capability of modeling the long-range spatial dependencies. Comprehensive experiments on three UAV challenging benchmarks demonstrate that the proposed TRTrack achieves superior UAV tracking performance in both precision and efficiency. Quantitative tests in real-world settings fully prove the effectiveness of our work.

2022

1. ICRA

   

   Ad 2 attack: Adaptive adversarial attack on real-time uav tracking

   Changhong Fu, Sihang Li, Xinnan Yuan, and 3 more authors

   In International Conference on Robotics and Automation, 2022

   Abs PDF

   Visual tracking is adopted to extensive unmanned aerial vehicle (UAV)-related applications, which leads to a highly demanding requirement on the robustness of UAV trackers. However, adding imperceptible perturbations can easily fool the tracker and cause tracking failures. This risk is often overlooked and rarely researched at present. Therefore, to help increase awareness of the potential risk and the robustness of UAV tracking, this work proposes a novel adaptive adversarial attack approach, i.e., Ad2Attack, against UAV object tracking. Specifically, adversarial examples are generated online during the resampling of the search patch image, which leads trackers to lose the target in the following frames. Ad2Attack is composed of a direct downsampling module and a super-resolution upsampling module with adaptive stages. A novel optimization function is proposed for balancing the imperceptibility and efficiency of the attack. Comprehensive experiments on several well-known benchmarks and real-world conditions show the effectiveness of our attack method, which dramatically reduces the performance of the most advanced Siamese trackers.