arXiv subjects

Jingyu Chen

Publications and source records attributed to Jingyu Chen.

12 recordsLinked to original sources

arXiv2026

See2Think: Do Multimodal Models Really Use Intermediate Visual States?

Multimodal large language models increasingly use sketches, annotations, tools, and intermediate images during reasoning, but it remains unclear whether they truly rely on these visual states. Existing benchmarks are limited both by task collections with narrow coverage or partially text-solvable samples and by evaluations that emphasize final answers without diagnosing how intermediate visual states are generated, rendered, and used. We introduce See2Think, a unified evaluation framework comprising See2ThinkBench and Visual Action-of-Thought (VAoT). See2ThinkBench contains 1,200 open-ended, visually dependent problems across 12 task categories spanning 2D structured, 3D scene, and real-world reasoning. VAoT records textual thoughts, visual actions, rendered states, and subsequent reasoning under four controlled inference settings. Evaluating representative proprietary and open-source multimodal models, we find that visual reasoning is strongly model- and environment-dependent, with no single setting consistently dominating across tasks. Process analysis further shows that models usually select relevant visual operations, while faithful rendering remains the clearest bottleneck and high feedback uptake does not necessarily translate into accuracy gains. Under task-relevant corrupted feedback, models exhibit behavioral dependence on visual states, with accuracy dropping by over 10 percentage points in controlled interventions.

cs.CV↗

arXiv2026

Adaptive Robust Estimator for Multi-Agent Reinforcement Learning

Multi-agent collaboration has emerged as a powerful paradigm for enhancing the reasoning capabilities of large language models, yet it suffers from interaction-level ambiguity that blurs generation, critique, and revision, making credit assignment across agents difficult. Moreover, policy optimization in this setting is vulnerable to heavy-tailed and noisy rewards, which can bias advantage estimation and trigger unstable or even divergent training. To address both issues, we propose a robust multi-agent reinforcement learning framework for collaborative reasoning, consisting of two components: Dual-Agent Answer-Critique-Rewrite (DACR) and an Adaptive Robust Estimator (ARE). DACR decomposes reasoning into a structured three-stage pipeline: answer, critique, and rewrite, while enabling explicit attribution of each agent's marginal contribution to its partner's performance. ARE provides robust estimation of batch experience means during multi-agent policy optimization. Across mathematical reasoning and embodied intelligence benchmarks, even under noisy rewards, our method consistently outperforms the baseline in both homogeneous and heterogeneous settings. These results indicate stronger robustness to reward noise and more stable training dynamics, effectively preventing optimization failures caused by noisy reward signals.

cs.AI↗

arXiv2025

Performance Analysis of Wireless-Powered Pinching Antenna Systems

Pinching antenna system (PAS) serves as a groundbreaking paradigm that enhances wireless communications by flexibly adjusting the position of pinching antenna (PA) and establishing a strong line-of-sight (LoS) link, thereby reducing the free-space path loss. This paper introduces the concept of wireless-powered PAS, and investigates the reliability of wireless-powered PAS to explore the advantages of PA in improving the performance of wireless-powered communication (WPC) system. In addition, we derive the closed-form expressions of outage probability and ergodic rate for the practical lossy waveguide case and ideal lossless waveguide case, respectively, and analyze the optimal deployment of waveguides and user to provide valuable insights for guiding their deployments. The results show that an increase in the absorption coefficient and in the dimensions of the user area leads to higher in-waveguide and free-space propagation losses, respectively, which in turn increase the outage probability and reduce the ergodic rate of the wireless-powered PAS. However, the performance of wireless-powered PAS is severely affected by the absorption coefficient and the waveguide length, e.g., under conditions of high absorption coefficient and long waveguide, the outage probability of wireless-powered PAS is even worse than that of traditional WPC system. While the ergodic rate of wireless-powered PAS is better than that of traditional WPC system under conditions of high absorption coefficient and long waveguide. Interestingly, the wireless-powered PAS has the optimal time allocation factor and optimal distance between power station (PS) and access point (AP) to minimize the outage probability or maximize the ergodic rate. Moreover, the system performance of PS and AP separated at the optimal distance between PS and AP is superior to that of PS and AP integrated into a hybrid access point.

eess.SP↗

arXiv2025

Towards Task-Oriented Flying: Framework, Infrastructure, and Principles

Deploying robot learning methods to aerial robots in unstructured environments remains both challenging and promising. While recent advances in deep reinforcement learning (DRL) have enabled end-to-end flight control, the field still lacks systematic design guidelines and a unified infrastructure to support reproducible training and real-world deployment. We present a task-oriented framework for end-to-end DRL in quadrotors that integrates design principles for complex task specification and reveals the interdependencies among simulated task definition, training design principles, and physical deployment. Our framework involves software infrastructure, hardware platforms, and open-source firmware to support a full-stack learning infrastructure and workflow. Extensive empirical results demonstrate robust flight and sim-to-real generalization under real-world disturbances. By reducing the entry barrier for deploying learning-based controllers on aerial robots, our work lays a practical foundation for advancing autonomous flight in dynamic and unstructured environments.

cs.RO↗

arXiv2025

Secure Wireless-Powered zeRIS Communications

This paper introduces the concept of wireless-powered zero-energy reconfigurable intelligent surface (zeRIS), and investigates a wireless-powered zeRIS aided communication system in terms of security, reliability and energy efficiency. In particular, we propose three new wireless-powered zeRIS modes: 1) in mode-I, N reconfigurable reflecting elements are adjusted to the optimal phase shift design of information user to maximize the reliability of the system; 2) in mode-II, N reconfigurable reflecting elements are adjusted to the optimal phase shift design of cooperative jamming user to maximize the security of the system; 3) in mode-III, N1 and N2 (N1+N2=N) reconfigurable reflecting elements are respectively adjusted to the optimal phase shift designs of information user and cooperative jamming user to balance the reliability and security of the system. Then, we propose three new metrics, i.e., joint outage probability (JOP), joint intercept probability (JIP), and secrecy energy efficiency (SEE), and analyze their closed-form expressions in three modes, respectively. The results show that under high transmission power, all the diversity gains of three modes are 1, and the JOPs of mode-I, mode-II and mode-III are improved by increasing the number of zeRIS elements, which are related to N2, N, and N^2_1, respectively. In addition, mode-I achieves the best JOP, while mode-II achieves the best JIP among three modes. We exploit two security-reliability trade-off (SRT) metrics, i.e., JOP versus JIP, and normalized joint intercept and outage probability (JIOP), to reveal the SRT performance of the proposed three modes. It is obtained that mode-II outperforms the other two modes in the JOP versus JIP, while mode-III and mode-II achieve the best performance of normalized JIOP at low and high transmission power, respectively.

eess.SP↗

arXiv2025

Enhancing Chest X-ray Classification through Knowledge Injection in Cross-Modality Learning

The integration of artificial intelligence in medical imaging has shown tremendous potential, yet the relationship between pre-trained knowledge and performance in cross-modality learning remains unclear. This study investigates how explicitly injecting medical knowledge into the learning process affects the performance of cross-modality classification, focusing on Chest X-ray (CXR) images. We introduce a novel Set Theory-based knowledge injection framework that generates captions for CXR images with controllable knowledge granularity. Using this framework, we fine-tune CLIP model on captions with varying levels of medical information. We evaluate the model's performance through zero-shot classification on the CheXpert dataset, a benchmark for CXR classification. Our results demonstrate that injecting fine-grained medical knowledge substantially improves classification accuracy, achieving 72.5\% compared to 49.9\% when using human-generated captions. This highlights the crucial role of domain-specific knowledge in medical cross-modality learning. Furthermore, we explore the influence of knowledge density and the use of domain-specific Large Language Models (LLMs) for caption generation, finding that denser knowledge and specialized LLMs contribute to enhanced performance. This research advances medical image analysis by demonstrating the effectiveness of knowledge injection for improving automated CXR classification, paving the way for more accurate and reliable diagnostic tools.

cs.CV↗

arXiv2024

A Thermomechanical Hybrid Incompressible Material Point Method

We present a novel hybrid incompressible flow/material point method solver for simulating the combustion of flammable solids. Our approach utilizes a sparse grid representation of solid materials in the material point method portion of the solver and a hybrid Eulerian/FLIP solver for the incompressible portion. We utilize these components to simulate the motion of heated air and particulate matter as they interact with flammable solids, causing combustion-related damage. We include a novel particle sampling strategy to increase Eulerian flow accuracy near regions of high temperature. We also support control of the flame front propagation speed and the rate of solid combustion in an artistically directable manner. Solid combustion is modeled with temperature-dependent elastoplastic constitutive modeling. We demonstrate the efficacy of our method on various real-world three-dimensional problems, including a burning match, incense sticks, and a wood log in a fireplace.

cs.GR↗

arXiv2023

Position-Based Nonlinear Gauss-Seidel for Quasistatic Hyperelasticity

Position based dynamics is a powerful technique for simulating a variety of materials. Its primary strength is its robustness when run with limited computational budget. We develop a novel approach to address problems with PBD for quasistatic hyperelastic materials. Even though PBD is based on the projection of static constraints, PBD is best suited for dynamic simulations. This is particularly relevant since the efficient creation of large data sets of plausible, but not necessarily accurate elastic equilibria is of increasing importance with the emergence of quasistatic neural networks. Furthermore, PBD projects one constraint at a time. We show that ignoring the effects of neighboring constraints limits its convergence and stability properties. Recent works have shown that PBD can be related to the Gauss-Seidel approximation of a Lagrange multiplier formulation of backward Euler time stepping, where each constraint is solved/projected independently of the others in an iterative fashion. We show that a position-based, rather than constraint-based nonlinear Gauss-Seidel approach solves these problems. Our approach retains the essential PBD feature of stable behavior with constrained computational budgets, but also allows for convergent behavior with expanded budgets. We demonstrate the efficacy of our method on a variety of representative hyperelastic problems and show that both successive over relaxation (SOR) and Chebyshev acceleration can be easily applied.

cs.GR↗

arXiv2023

UniFusion: Unified Multi-view Fusion Transformer for Spatial-Temporal Representation in Bird's-Eye-View

Bird's eye view (BEV) representation is a new perception formulation for autonomous driving, which is based on spatial fusion. Further, temporal fusion is also introduced in BEV representation and gains great success. In this work, we propose a new method that unifies both spatial and temporal fusion and merges them into a unified mathematical formulation. The unified fusion could not only provide a new perspective on BEV fusion but also brings new capabilities. With the proposed unified spatial-temporal fusion, our method could support long-range fusion, which is hard to achieve in conventional BEV methods. Moreover, the BEV fusion in our work is temporal-adaptive and the weights of temporal fusion are learnable. In contrast, conventional methods mainly use fixed and equal weights for temporal fusion. Besides, the proposed unified fusion could avoid information lost in conventional BEV fusion methods and make full use of features. Extensive experiments and ablation studies on the NuScenes dataset show the effectiveness of the proposed method and our method gains the state-of-the-art performance in the map segmentation task.

cs.CV↗

arXiv2022

A Deep Conjugate Direction Method for Iteratively Solving Linear Systems

We present a novel deep learning approach to approximate the solution of large, sparse, symmetric, positive-definite linear systems of equations. These systems arise from many problems in applied science, e.g., in numerical methods for partial differential equations. Algorithms for approximating the solution to these systems are often the bottleneck in problems that require their solution, particularly for modern applications that require many millions of unknowns. Indeed, numerical linear algebra techniques have been investigated for many decades to alleviate this computational burden. Recently, data-driven techniques have also shown promise for these problems. Motivated by the conjugate gradients algorithm that iteratively selects search directions for minimizing the matrix norm of the approximation error, we design an approach that utilizes a deep neural network to accelerate convergence via data-driven improvement of the search directions. Our method leverages a carefully chosen convolutional network to approximate the action of the inverse of the linear operator up to an arbitrary constant. We train the network using unsupervised learning with a loss function equal to the $L^2$ difference between an input and the system matrix times the network evaluation, where the unspecified constant in the approximate inverse is accounted for. We demonstrate the efficacy of our approach on spatially discretized Poisson equations with millions of degrees of freedom arising in computational fluid dynamics applications. Unlike state-of-the-art learning approaches, our algorithm is capable of reducing the linear system residual to a given tolerance in a small number of iterations, independent of the problem size. Moreover, our method generalizes effectively to various systems beyond those encountered during training.

cs.LG↗

arXiv2022

PieTrack: An MOT solution based on synthetic data training and self-supervised domain adaptation

In order to cope with the increasing demand for labeling data and privacy issues with human detection, synthetic data has been used as a substitute and showing promising results in human detection and tracking tasks. We participate in the 7th Workshop on Benchmarking Multi-Target Tracking (BMTT), themed on "How Far Can Synthetic Data Take us"? Our solution, PieTrack, is developed based on synthetic data without using any pre-trained weights. We propose a self-supervised domain adaptation method that enables mitigating the domain shift issue between the synthetic (e.g., MOTSynth) and real data (e.g., MOT17) without involving extra human labels. By leveraging the proposed multi-scale ensemble inference, we achieved a final HOTA score of 58.7 on the MOT17 testing set, ranked third place in the challenge.

cs.CV↗

arXiv2021

A Momentum-Conserving Implicit Material Point Method for Surface Energies with Spatial Gradients

We present a novel Material Point Method (MPM) discretization of surface tension forces that arise from spatially varying surface energies. These variations typically arise from surface energy dependence on temperature and/or concentration. Furthermore, since the surface energy is an interfacial property depending on the types of materials on either side of an interface, spatial variation is required for modeling the contact angle at the triple junction between a liquid, solid and surrounding air. Our discretization is based on the surface energy itself, rather than on the associated traction condition most commonly used for discretization with particle methods. Our energy based approach automatically captures surface gradients without the explicit need to resolve them as in traction condition based approaches. We include an implicit discretization of thermomechanical material coupling with a novel particle-based enforcement of Robin boundary conditions associated with convective heating. Lastly, we design a particle resampling approach needed to achieve perfect conservation of linear and angular momentum with AffineParticle-In-Cell (APIC) [Jiang et al. 2015]. We show that our approach enables implicit time stepping for complex behaviors like the Marangoni effect and hydrophobicity/hydrophilicity. We demonstrate the robustness and utility of our method by simulating materials that exhibit highly diverse degrees of surface tension and thermomechanical effects, such as water, wine and wax.

cs.GR↗