arXiv subjects

Zhongyu Wang

Publications and source records attributed to Zhongyu Wang.

10 recordsLinked to original sources

arXiv2026

GraMRAG: Orchestrating Multi-Agent Multi-Step Reasoning via Graph Memory with Reinforcement Learning

Although existing multi-agent Retrieval-Augmented Generation (RAG) systems have demonstrated promise on complex multimodal reasoning tasks, they remain fundamentally limited in reasoning depth and memory structure, suffering from inadequate retrieval and state blindness when answering knowledge-intensive questions. To address these limitations, we propose GraMRAG, a graph memory-guided multi-agent RAG framework that integrates a dynamic multimodal memory graph to enable stable, multi-step multimodal reasoning. We introduce a vision-text bridged reasoning paradigm that unifies multi-scale entity cropping with a ReAct-style visual toolchain, enhancing the long-horizon cross-modal reasoning capability. We further construct a multimodal memory graph that formalizes agent reasoning as a dynamic directed acyclic graph (DAG), explicitly modeling action-observation dependencies to mitigate state blindness and suppress redundant retrieval. Moreover, we propose Topology-Aware Policy Optimization (TAPO) that leverages graph topology for critical path identification and targeted node pruning, enabling fine-grained credit assignment across multi-step reasoning trajectories. Extensive experiments on challenging multimodal benchmarks demonstrate that our approach consistently outperforms existing baselines and achieves state-of-the-art performance on complex long-horizon reasoning tasks.

cs.CL↗

arXiv2026

SymphoMotion: Joint Control of Camera Motion and Object Dynamics for Coherent Video Generation

Controlling both camera motion and object dynamics is essential for coherent and expressive video generation, yet current methods typically handle only one motion type or rely on ambiguous 2D cues that entangle camera-induced parallax with true object movement. We present SymphoMotion, a unified motion-control framework that jointly governs camera trajectories and object dynamics within a single model. SymphoMotion features a Camera Trajectory Control mechanism that integrates explicit camera paths with geometry-aware cues to ensure stable, structurally consistent viewpoint transitions, and an Object Dynamics Control mechanism that combines 2D visual guidance with 3D trajectory embeddings to enable depth-aware, spatially coherent object manipulation. To support large-scale training and evaluation, we further construct RealCOD-25K, a comprehensive real-world dataset containing paired camera poses and object-level 3D trajectories across diverse indoor and outdoor scenes, addressing a key data gap in unified motion control. Extensive experiments and user studies show that SymphoMotion significantly outperforms existing methods in visual fidelity, camera controllability, and object-motion accuracy, establishing a new benchmark for unified motion control in video generation. Codes and data are publicly available at https://grenoble-zhang.github.io/SymphoMotion/.

cs.CV↗

arXiv2026

AdaTooler-V: Adaptive Tool-Use for Images and Videos

Recent advances have shown that multimodal large language models (MLLMs) benefit from multimodal interleaved chain-of-thought (CoT) with vision tool interactions. However, existing open-source models often exhibit blind tool-use reasoning patterns, invoking vision tools even when they are unnecessary, which significantly increases inference overhead and degrades model performance. To this end, we propose AdaTooler-V, an MLLM that performs adaptive tool-use by determining whether a visual problem truly requires tools. First, we introduce AT-GRPO, a reinforcement learning algorithm that adaptively adjusts reward scales based on the Tool Benefit Score of each sample, encouraging the model to invoke tools only when they provide genuine improvements. Moreover, we construct two datasets to support training: AdaTooler-V-CoT-100k for SFT cold start and AdaTooler-V-300k for RL with verifiable rewards across single-image, multi-image, and video data. Experiments across twelve benchmarks demonstrate the strong reasoning capability of AdaTooler-V, outperforming existing methods in diverse visual reasoning tasks. Notably, AdaTooler-V-7B achieves an accuracy of 89.8\% on the high-resolution benchmark V*, surpassing the commercial proprietary model GPT-4o and Gemini 1.5 Pro. All code, models, and data are released.

cs.CV↗

arXiv2026

"Range as a Key" is the Key! Fast and Compact Cloud Block Store Index with RASK

In cloud block store, indexing is on the critical path of I/O operations and typically resides in memory. With the scaling of users and the emergence of denser storage media, the index has become a primary memory consumer, causing memory strain. Our extensive analysis of production traces reveals that write requests exhibit a strong tendency to target continuous block ranges in cloud storage systems. Thus, compared to current per-block indexing, our insight is that we should directly index block ranges (i.e., range-as-a-key) to save memory. In this paper, we propose RASK, a memory-efficient and high-performance tree-structured index that natively indexes ranges. While range-as-a-key offers the potential to save memory and improve performance, realizing this idea is challenging due to the range overlap and range fragmentation issues. To handle range overlap efficiently, RASK introduces the log-structured leaf, combined with range-tailored search and garbage collection. To reduce range fragmentation, RASK employs range-aware split and merge mechanisms. Our evaluations on four production traces show that RASK reduces memory footprint by up to 98.9% and increases throughput by up to 31.0x compared to ten state-of-the-art indexes.

cs.OS↗

arXiv2025

Macro-from-Micro Planning for High-Quality and Parallelized Autoregressive Long Video Generation

Current autoregressive diffusion models excel at video generation but are generally limited to short temporal durations. Our theoretical analysis indicates that the autoregressive modeling typically suffers from temporal drift caused by error accumulation and hinders parallelization in long video synthesis. To address these limitations, we propose a novel planning-then-populating framework centered on Macro-from-Micro Planning (MMPL) for long video generation. MMPL sketches a global storyline for the entire video through two hierarchical stages: Micro Planning and Macro Planning. Specifically, Micro Planning predicts a sparse set of future keyframes within each short video segment, offering motion and appearance priors to guide high-quality video segment generation. Macro Planning extends the in-segment keyframes planning across the entire video through an autoregressive chain of micro plans, ensuring long-term consistency across video segments. Subsequently, MMPL-based Content Populating generates all intermediate frames in parallel across segments, enabling efficient parallelization of autoregressive generation. The parallelization is further optimized by Adaptive Workload Scheduling for balanced GPU execution and accelerated autoregressive video generation. Extensive experiments confirm that our method outperforms existing long video generation models in quality and stability. Generated videos and comparison results are in our project page.

cs.CV↗

arXiv2023

Freshness or Accuracy, Why Not Both? Addressing Delayed Feedback via Dynamic Graph Neural Networks

The delayed feedback problem is one of the most pressing challenges in predicting the conversion rate since users' conversions are always delayed in online commercial systems. Although new data are beneficial for continuous training, without complete feedback information, i.e., conversion labels, training algorithms may suffer from overwhelming fake negatives. Existing methods tend to use multitask learning or design data pipelines to solve the delayed feedback problem. However, these methods have a trade-off between data freshness and label accuracy. In this paper, we propose Delayed Feedback Modeling by Dynamic Graph Neural Network (DGDFEM). It includes three stages, i.e., preparing a data pipeline, building a dynamic graph, and training a CVR prediction model. In the model training, we propose a novel graph convolutional method named HLGCN, which leverages both high-pass and low-pass filters to deal with conversion and non-conversion relationships. The proposed method achieves both data freshness and label accuracy. We conduct extensive experiments on three industry datasets, which validate the consistent superiority of our method.

cs.LG↗

arXiv2023

Decentralized Graph Neural Network for Privacy-Preserving Recommendation

Building a graph neural network (GNN)-based recommender system without violating user privacy proves challenging. Existing methods can be divided into federated GNNs and decentralized GNNs. But both methods have undesirable effects, i.e., low communication efficiency and privacy leakage. This paper proposes DGREC, a novel decentralized GNN for privacy-preserving recommendations, where users can choose to publicize their interactions. It includes three stages, i.e., graph construction, local gradient calculation, and global gradient passing. The first stage builds a local inner-item hypergraph for each user and a global inter-user graph. The second stage models user preference and calculates gradients on each local device. The third stage designs a local differential privacy mechanism named secure gradient-sharing, which proves strong privacy-preserving of users' private data. We conduct extensive experiments on three public datasets to validate the consistent superiority of our framework.

cs.IR↗

arXiv2022

Placement and Resource Allocation of Wireless-Powered Multiantenna UAV for Energy-Efficient Multiuser NOMA

This paper investigates a new downlink nonorthogonal multiple access (NOMA) system, where a multiantenna unmanned aerial vehicle (UAV) is powered by wireless power transfer (WPT) and serves as the base station for multiple pairs of ground users (GUs) running NOMA in each pair. An energy efficiency (EE) maximization problem is formulated to jointly optimize the WPT time and the placement for the UAV, and the allocation of the UAV's transmit power between different NOMA user pairs and within each pair. To efficiently solve this nonconvex problem, we decompose the problem into three subproblems using block coordinate descent. For the subproblem of intra-pair power allocation within each NOMA user pair, we construct a supermodular game with confirmed convergence to a Nash equilibrium. Given the intra-pair power allocation, successive convex approximation is applied to convexify and solve the subproblem of WPT time allocation and inter-pair power allocation between the user pairs. Finally, we solve the subproblem of UAV placement by using the Lagrange multiplier method. Simulations show that our approach can substantially outperform its alternatives that do not use NOMA and WPT techniques or that do not optimize the UAV location.

cs.IT↗

arXiv2021

Energy-Efficient Resource Allocation in Massive MIMO-NOMA Networks with Wireless Power Transfer: A Distributed ADMM Approach

In multicell massive multiple-input multiple-output (MIMO) non-orthogonal multiple access (NOMA) networks, base stations (BSs) with multiple antennas deliver their radio frequency energy in the downlink, and Internet-of-Things (IoT) devices use their harvested energy to support uplink data transmission. This paper investigates the energy efficiency (EE) problem for multicell massive MIMO NOMA networks with wireless power transfer (WPT). To maximize the EE of the network, we propose a novel joint power, time, antenna selection, and subcarrier resource allocation scheme, which can properly allocate the time for energy harvesting and data transmission. Both perfect and imperfect channel state information (CSI) are considered, and their corresponding EE performance is analyzed. Under quality-of-service (QoS) requirements, an EE maximization problem is formulated, which is non-trivial due to non-convexity. We first adopt nonlinear fraction programming methods to convert the problem to be convex, and then, develop a distributed alternating direction method of multipliers (ADMM)- based approach to solve the problem. Simulation results demonstrate that compared to alternative methods, the proposed algorithm can converge quickly within fewer iterations, and can achieve better EE performance.

cs.IT↗

arXiv2018

Ultra-sensitive Experimental Detection of Chiral Meso-structures by Orbital Angular Momentum of Light

Circular dichroism (CD) induced by spin angular momentum of light is vital to investigate the chirality of microscopic objects such as molecules, proteins and metamaterials. However, orbital angular momentum (OAM) of light failed to directly implement an interaction with chiral molecules, which remains poorly understood. Even with the enhancement of unique plasmonic nanoparticle aggregates, the helical dichroism (HD) is still at a small level of ~0.6%. Here, we experimentally report the direct observation of a giant HD at the level of ~120% by matching the scales of chiral microstructures and OAM beams. Our results reveal a strong interaction between OAM beams and chiral microstructures in terms of their distinct reflectance, resulting in sign-opposite HD values to distinguish their chirality. Such unique HD phenomenon is investigated experimentally under various diameters and helical pitches of chiral structures, showing the robust performance. This HD technique can widely extend the responding areas of chiral spectroscopy in biology, material science and nanotechnology.

physics.optics↗