arXiv subjects

Yushan Zhang

Publications and source records attributed to Yushan Zhang.

17 recordsLinked to original sources

arXiv2026

GenHOI: Contact-Aware Humanoid-Object Interaction by Imitating Generated Videos without Task-Specific Training

Humanoid-Object Interaction (HOI) is a fundamental capability for humanoid robots, yet it remains challenging due to the tight coupling between dynamic balance and stable interaction with diverse objects. Existing methods often require time-consuming task-specific policy training or rely on rigid trajectory replay, which limits their ability to accommodate novel interaction scenarios. In this work, we present \textit{GenHOI}, a simple yet effective framework that enables humanoid robots to perform diverse object-interaction tasks in a zero-shot manner by directly imitating a single generated video, without task-specific training or physical demonstration data. GenHOI first reconstructs the robot-object scene in simulation and renders a first-frame image, which, together with the language command, conditions the synthesis of a task-oriented interaction video. The generated video is then analyzed to identify interaction-relevant contact events and estimate hand-object contact regions, which are encoded as object-centric geometric constraints that convert visual interaction cues into physically grounded optimization priors. Guided by these priors, the reference motion recovered from the video is refined and smoothed to resolve the scale ambiguity inherent in 2D video generation, while adapting a single reference trajectory to unseen robot-object relative poses. The optimized trajectory is finally executed by a closed-loop tracking controller. We validate the proposed framework in extensive simulation and real-world experiments across diverse object-interaction tasks, including box grasping, asymmetric bimanual chair carrying, table lifting from below, and cylindrical-object enveloping.

cs.RO↗

arXiv2026

Graph-Regularized Deep Learning for EEG-Based Emotion Recognition with Psychologically-Grounded Label Structure

EEG-based emotion recognition is critical for mental health monitoring and affective brain-computer interfaces, yet existing deep learning approaches often treat emotion classes as isolated labels, ignoring their psychological interdependencies. We propose a graph-regularized learning framework that conceptualizes emotions as nodes in a graph where edges encode proximity based on dimensional emotion theories. We adapt three complementary regularization strategies--Graph Label Smoothing (intuitive soft labeling), Commuting distance on graph via Graph Laplacian (spectral graph theory), and Sliced Wasserstein Distance (optimal transport on graph)--ordered by increasing computational complexity. These strategies penalize model predictions that deviate from the established emotion topology. Our framework is evaluated across three representative backbone architectures: AudioTransformer (pure transformer), Conformer (CNN-transformer hybrid), and DCGNN (causal graph neural network), demonstrating architecture-agnostic benefits. Experiments on SEED-IV (4 classes) and SEED-V (5 classes) datasets show consistent improvements: best case up to +5.42% accuracy and 39% reduction in psychologically implausible misclassifications. Ultimately, our framework help raise the upper bound of performance achievable with standard approaches. Code will be released.

cs.LG↗

arXiv2026

RoMa v2: Harder Better Faster Denser Feature Matching

Dense feature matching aims to estimate all correspondences between two images of a 3D scene and has recently been established as the gold standard due to its high accuracy and robustness. However, existing dense matchers still fail or perform poorly for many hard real-world scenarios, and high-precision models are often slow, limiting their applicability. In this paper, we attack these weaknesses on a wide front through a series of systematic improvements that together yield a significantly better model. In particular, we construct a novel matching architecture and loss, which, combined with a curated diverse training distribution, enables our model to solve many complex matching tasks. We further make training faster through a decoupled two-stage matching-then-refinement pipeline, and at the same time, significantly reduce refinement memory usage through a custom CUDA kernel. Finally, we leverage the recent DINOv3 foundation model along with multiple other insights to make the model more robust and unbiased. In our extensive set of experiments, we show that the resulting novel matcher sets a new state-of-the-art, being significantly more accurate than its predecessors. Code is available at https://github.com/Parskatt/romav2

cs.CV↗

arXiv2026

TeFlow: Enabling Multi-frame Supervision for Self-Supervised Feed-forward Scene Flow Estimation

Self-supervised feed-forward methods for scene flow estimation offer real-time efficiency, but their supervision from two-frame point correspondences is unreliable and often breaks down under occlusions. Multi-frame supervision has the potential to provide more stable guidance by incorporating motion cues from past frames, yet naive extensions of two-frame objectives are ineffective because point correspondences vary abruptly across frames, producing inconsistent signals. In the paper, we present TeFlow, enabling multi-frame supervision for feed-forward models by mining temporally consistent supervision. TeFlow introduces a temporal ensembling strategy that forms reliable supervisory signals by aggregating the most temporally consistent motion cues from a candidate pool built across multiple frames. Extensive evaluations demonstrate that TeFlow establishes a new state-of-the-art for self-supervised feed-forward methods, achieving performance gains of up to 33\% on the challenging Argoverse 2 and nuScenes datasets. Our method performs on par with leading optimization-based methods, yet speeds up 150 times. The code is open-sourced at https://github.com/Kin-Zhang/TeFlow along with trained model weights.

cs.CV↗

arXiv2026

ALORE: Autonomous Large-Object Rearrangement with a Legged Manipulator

Endowing robots with the ability to rearrange various large and heavy objects, such as furniture, can substantially alleviate human workload. However, this task is extremely challenging due to the need to interact with diverse objects and efficiently rearrange multiple objects in complex environments while ensuring collision-free loco-manipulation. In this work, we present ALORE, an autonomous large-object rearrangement system for a legged manipulator that can rearrange various large objects across diverse scenarios. The proposed system is characterized by three main features: (i) a hierarchical reinforcement learning training pipeline for multi-object environment learning, where a high-level object velocity controller is trained on top of a low-level whole-body controller to achieve efficient and stable joint learning across multiple objects; (ii) two key modules, a unified interaction configuration representation and an object velocity estimator, that allow a single policy to regulate planar velocity of diverse objects accurately; and (iii) a task-and-motion planning framework that jointly optimizes object visitation order and object-to-target assignment, improving task efficiency while enabling online replanning. Comparisons against strong baselines show consistent superiority in policy generalization, object-velocity tracking accuracy, and multi-object rearrangement efficiency. Key modules are systematically evaluated, and extensive simulations and real-world experiments are conducted to validate the robustness and effectiveness of the entire system, which successfully completes 8 continuous loops to rearrange 32 chairs over nearly 40 minutes without a single failure, and executes long-distance autonomous rearrangement over an approximately 40 m route. The open-source packages are available at https://zhihaibi.github.io/Alore/.

cs.RO↗

arXiv2025

DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Method

Previous dominant methods for scene flow estimation focus mainly on input from two consecutive frames, neglecting valuable information in the temporal domain. While recent trends shift towards multi-frame reasoning, they suffer from rapidly escalating computational costs as the number of frames grows. To leverage temporal information more efficiently, we propose DeltaFlow ($Δ$Flow), a lightweight 3D framework that captures motion cues via a $Δ$ scheme, extracting temporal features with minimal computational cost, regardless of the number of frames. Additionally, scene flow estimation faces challenges such as imbalanced object class distributions and motion inconsistency. To tackle these issues, we introduce a Category-Balanced Loss to enhance learning across underrepresented classes and an Instance Consistency Loss to enforce coherent object motion, improving flow accuracy. Extensive evaluations on the Argoverse 2, Waymo and nuScenes datasets show that $Δ$Flow achieves state-of-the-art performance with up to 22% lower error and $2\times$ faster inference compared to the next-best multi-frame supervised method, while also demonstrating a strong cross-domain generalization ability. The code is open-sourced at https://github.com/Kin-Zhang/DeltaFlow along with trained model weights.

cs.CV↗

arXiv2025

Visual Jigsaw Post-Training Improves MLLMs

Reinforcement learning based post-training has recently emerged as a powerful paradigm for enhancing the alignment and reasoning capabilities of multimodal large language models (MLLMs). While vision-centric post-training is crucial for enhancing MLLMs' intrinsic understanding of visual signals, current post-training paradigms are predominantly text-centric, where dense visual inputs are only leveraged to extract sparse cues for text-based reasoning. There exist a few approaches in this direction, however, they often still rely on text as an intermediate mediator or introduce additional visual generative designs. In this work, we introduce Visual Jigsaw, a generic self-supervised post-training framework designed to strengthen visual understanding in MLLMs. Visual Jigsaw is formulated as a general ordering task: visual inputs are partitioned, shuffled, and the model must reconstruct the visual information by producing the correct permutation in natural language. This naturally aligns with reinforcement learning from verifiable rewards (RLVR), requires no additional visual generative components, and derives its supervisory signal automatically without any annotations. We instantiate Visual Jigsaw across three visual modalities, including images, videos, and 3D data. Extensive experiments demonstrate substantial improvements in fine-grained perception, temporal reasoning, and 3D spatial understanding. Our findings highlight the potential of self-supervised vision-centric tasks in post-training MLLMs and aim to inspire further research on vision-centric pretext designs. Project Page: https://penghao-wu.github.io/visual_jigsaw/

cs.CV↗

arXiv2025

Running-time Analysis of ($μ+λ$) Evolutionary Combinatorial Optimization Based on Multiple-gain Estimation

The running-time analysis of evolutionary combinatorial optimization is a fundamental topic in evolutionary computation. However, theoretical results regarding the $(μ+λ)$ evolutionary algorithm (EA) for combinatorial optimization problems remain relatively scarce compared to those for simple pseudo-Boolean problems. This paper proposes a multiple-gain model to analyze the running time of EAs for combinatorial optimization problems. The proposed model is an improved version of the average gain model, which is a fitness-difference drift approach under the sigma-algebra condition to estimate the running time of evolutionary numerical optimization. The improvement yields a framework for estimating the expected first hitting time of a stochastic process in both average-case and worst-case scenarios. It also introduces novel running-time results of evolutionary combinatorial optimization, including two tighter time complexity upper bounds than the known results in the case of ($μ+λ$) EA for the knapsack problem with favorably correlated weights, a closed-form expression of time complexity upper bound in the case of ($μ+λ$) EA for general $k$-MAX-SAT problems and a tighter time complexity upper bounds than the known results in the case of ($μ+λ$) EA for the traveling salesperson problem. Experimental results indicate that the practical running time aligns with the theoretical results, verifying that the multiple-gain model is an effective tool for running-time analysis of ($μ+λ$) EA for combinatorial optimization problems.

cs.NE↗

arXiv2025

Zero-Shot 4D Lidar Panoptic Segmentation

Zero-shot 4D segmentation and recognition of arbitrary objects in Lidar is crucial for embodied navigation, with applications ranging from streaming perception to semantic mapping and localization. However, the primary challenge in advancing research and developing generalized, versatile methods for spatio-temporal scene understanding in Lidar lies in the scarcity of datasets that provide the necessary diversity and scale of annotations.To overcome these challenges, we propose SAL-4D (Segment Anything in Lidar--4D), a method that utilizes multi-modal robotic sensor setups as a bridge to distill recent developments in Video Object Segmentation (VOS) in conjunction with off-the-shelf Vision-Language foundation models to Lidar. We utilize VOS models to pseudo-label tracklets in short video sequences, annotate these tracklets with sequence-level CLIP tokens, and lift them to the 4D Lidar space using calibrated multi-modal sensory setups to distill them to our SAL-4D model. Due to temporal consistent predictions, we outperform prior art in 3D Zero-Shot Lidar Panoptic Segmentation (LPS) over $5$ PQ, and unlock Zero-Shot 4D-LPS.

cs.CV↗

arXiv2025

Multiple-gain Estimation for Running Time of Evolutionary Combinatorial Optimization

The running-time analysis of evolutionary combinatorial optimization is a fundamental topic in evolutionary computation. Its current research mainly focuses on specific algorithms for simplified problems due to the challenge posed by fluctuating fitness values. This paper proposes a multiple-gain model to estimate the fitness trend of population during iterations. The proposed model is an improved version of the average gain model, which is the approach to estimate the running time of evolutionary algorithms for numerical optimization. The improvement yields novel results of evolutionary combinatorial optimization, including a briefer proof for the time complexity upper bound in the case of (1+1) EA for the Onemax problem, two tighter time complexity upper bounds than the known results in the case of (1+$λ$) EA for the knapsack problem with favorably correlated weights and a closed-form expression of time complexity upper bound in the case of (1+$λ$) EA for general $k$-MAX-SAT problems. The results indicate that the practical running time aligns with the theoretical results, verifying that the multiple-gain model is more general for running-time analysis of evolutionary combinatorial optimization than state-of-the-art methods.

cs.NE↗

arXiv2024

DiffSF: Diffusion Models for Scene Flow Estimation

Scene flow estimation is an essential ingredient for a variety of real-world applications, especially for autonomous agents, such as self-driving cars and robots. While recent scene flow estimation approaches achieve a reasonable accuracy, their applicability to real-world systems additionally benefits from a reliability measure. Aiming at improving accuracy while additionally providing an estimate for uncertainty, we propose DiffSF that combines transformer-based scene flow estimation with denoising diffusion models. In the diffusion process, the ground truth scene flow vector field is gradually perturbed by adding Gaussian noise. In the reverse process, starting from randomly sampled Gaussian noise, the scene flow vector field prediction is recovered by conditioning on a source and a target point cloud. We show that the diffusion process greatly increases the robustness of predictions compared to prior approaches resulting in state-of-the-art performance on standard scene flow estimation benchmarks. Moreover, by sampling multiple times with different initial states, the denoising process predicts multiple hypotheses, which enables measuring the output uncertainty, allowing our approach to detect a majority of the inaccurate predictions. The code is available at https://github.com/ZhangYushan3/DiffSF.

cs.CV↗

arXiv2024

High-fidelity Pseudo-labels for Boosting Weakly-Supervised Segmentation

Image-level weakly-supervised semantic segmentation (WSSS) reduces the usually vast data annotation cost by surrogate segmentation masks during training. The typical approach involves training an image classification network using global average pooling (GAP) on convolutional feature maps. This enables the estimation of object locations based on class activation maps (CAMs), which identify the importance of image regions. The CAMs are then used to generate pseudo-labels, in the form of segmentation masks, to supervise a segmentation model in the absence of pixel-level ground truth. Our work is based on two techniques for improving CAMs; importance sampling, which is a substitute for GAP, and the feature similarity loss, which utilizes a heuristic that object contours almost always align with color edges in images. However, both are based on the multinomial posterior with softmax, and implicitly assume that classes are mutually exclusive, which turns out suboptimal in our experiments. Thus, we reformulate both techniques based on binomial posteriors of multiple independent binary problems. This has two benefits; their performance is improved and they become more general, resulting in an add-on method that can boost virtually any WSSS method. This is demonstrated on a wide variety of baselines on the PASCAL VOC dataset, improving the region similarity and contour quality of all implemented state-of-the-art methods. Experiments on the MS COCO dataset further show that our proposed add-on is well-suited for large-scale settings. Our code implementation is available at https://github.com/arvijj/hfpl.

cs.CV↗

arXiv2023

GMSF: Global Matching Scene Flow

We tackle the task of scene flow estimation from point clouds. Given a source and a target point cloud, the objective is to estimate a translation from each point in the source point cloud to the target, resulting in a 3D motion vector field. Previous dominant scene flow estimation methods require complicated coarse-to-fine or recurrent architectures as a multi-stage refinement. In contrast, we propose a significantly simpler single-scale one-shot global matching to address the problem. Our key finding is that reliable feature similarity between point pairs is essential and sufficient to estimate accurate scene flow. We thus propose to decompose the feature extraction step via a hybrid local-global-cross transformer architecture which is crucial to accurate and robust feature representations. Extensive experiments show that the proposed Global Matching Scene Flow (GMSF) sets a new state-of-the-art on multiple scene flow estimation benchmarks. On FlyingThings3D, with the presence of occlusion points, GMSF reduces the outlier percentage from the previous best performance of 27.4% to 5.6%. On KITTI Scene Flow, without any fine-tuning, our proposed method shows state-of-the-art performance. On the Waymo-Open dataset, the proposed method outperforms previous methods by a large margin. The code is available at https://github.com/ZhangYushan3/GMSF.

cs.CV↗

arXiv2023

Flow-guided Semi-supervised Video Object Segmentation

We propose an optical flow-guided approach for semi-supervised video object segmentation. Optical flow is usually exploited as additional guidance information in unsupervised video object segmentation. However, its relevance in semi-supervised video object segmentation has not been fully explored. In this work, we follow an encoder-decoder approach to address the segmentation task. A model to extract the combined information from optical flow and the image is proposed, which is then used as input to the target model and the decoder network. Unlike previous methods where concatenation is used to integrate information from image data and optical flow, a simple yet effective attention mechanism is exploited in our work. Experiments on DAVIS 2017 and YouTube-VOS 2019 show that by integrating the information extracted from optical flow into the original image branch results in a strong performance gain and our method achieves state-of-the-art performance.

cs.CV↗

arXiv2021

Duplicate-sensitivity Guided Transformation Synthesis for DBMS Correctness Bug Detection

Database Management System (DBMS) plays a core role in modern software from mobile apps to online banking. It is critical that DBMS should provide correct data to all applications. When the DBMS returns incorrect data, a correctness bug is triggered. Current production-level DBMSs still suffer from insufficient testing due to the limited hand-written test cases. Recently several works proposed to automatically generate many test cases with query transformation, a process of generating an equivalent query pair and testing a DBMS by checking whether the system returns the same result set for both queries. However, all of them still heavily rely on manual work to provide a transformation which largely confines their exploration of the valid input query space. This paper introduces duplicate-sensitivity guided transformation synthesis which automatically finds new transformations by first synthesizing many candidates then filtering the nonequivalent ones. Our automated synthesis is achieved by mutating a query while keeping its duplicate sensitivity, which is a necessary condition for query equivalence. After candidate synthesis, we keep the mutant query which is equivalent to the given one by using a query equivalent checker. Furthermore, we have implemented our idea in a tool Eqsql and used it to test the production-level DBMSs. In two months, we detected in total 30 newly confirmed and unique bugs in MySQL, TiDB and CynosDB.

cs.SE↗

arXiv2020

Ensemble model aggregation using a computationally lightweight machine-learning model to forecast ocean waves

This study investigated an approach to improve the accuracy of computationally lightweight surrogate models by updating forecasts based on historical accuracy relative to sparse observation data. Using a lightweight, ocean-wave forecasting model, we created a large number of model ensembles, with perturbed inputs, for a two-year study period. Forecasts were aggregated using a machine-learning algorithm that combined forecasts from multiple, independent models into a single "best-estimate" prediction of the true state. The framework was applied to a case-study site in Monterey Bay, California. A~learning-aggregation technique used historical observations and model forecasts to calculate a weight for each ensemble member. Weighted ensemble predictions were compared to measured wave conditions to evaluate performance against present state-of-the-art. Finally, we discussed how this framework, which integrates ensemble aggregations and surrogate models, can be used to improve forecasting systems and further enable scientific process studies.

physics.ao-ph↗

arXiv2017

A Machine Learning Framework to Forecast Wave Conditions

A~machine learning framework is developed to estimate ocean-wave conditions. By supervised training of machine learning models on many thousands of iterations of a physics-based wave model, accurate representations of significant wave heights and period can be used to predict ocean conditions. A model of Monterey Bay was used as the example test site; it was forced by measured wave conditions, ocean-current nowcasts, and reported winds. These input data along with model outputs of spatially variable wave heights and characteristic period were aggregated into supervised learning training and test data sets, which were supplied to machine learning models. These machine learning models replicated wave heights with a root-mean-squared error of 9cm and correctly identify over 90% of the characteristic periods for the test-data sets. Impressively, transforming model inputs to outputs through matrix operations requires only a fraction (<1/1,000) of the computation time compared to forecasting with the physics-based model.

physics.ao-ph↗