arXiv subjects

Ziyuan Guo

Publications and source records attributed to Ziyuan Guo.

11 recordsLinked to original sources

arXiv2026

FireRed-Image-Edit-1.0 Technical Report

We present FireRed-Image-Edit, a diffusion transformer for instruction-based image editing that achieves state-of-the-art performance through systematic optimization of data curation, training methodology, and evaluation design. We construct a 1.6B-sample training corpus, comprising 900M text-to-image and 700M image editing pairs from diverse sources. After rigorous cleaning, stratification, auto-labeling, and two-stage filtering, we retain over 100M high-quality samples balanced between generation and editing, ensuring strong semantic coverage and instruction alignment. Our multi-stage training pipeline progressively builds editing capability via pre-training, supervised fine-tuning, and reinforcement learning. To improve data efficiency, we introduce a Multi-Condition Aware Bucket Sampler for variable-resolution batching and Stochastic Instruction Alignment with dynamic prompt re-indexing. To stabilize optimization and enhance controllability, we propose Asymmetric Gradient Optimization for DPO, DiffusionNFT with layout-aware OCR rewards for text editing, and a differentiable Consistency Loss for identity preservation. We further establish REDEdit-Bench, a comprehensive benchmark spanning 15 editing categories, including newly introduced beautification and low-level enhancement tasks. Extensive experiments on REDEdit-Bench and public benchmarks (ImgEdit and GEdit) demonstrate competitive or superior performance against both open-source and proprietary systems. To support future research, our code, models, and benchmark suite are publicly available at https://github.com/FireRedTeam/FireRed-Image-Edit/ .

cs.CV↗

arXiv2026

Distributed Optimal Consensus of Nonlinear Multi-Agent Systems

In this paper, the optimal consensus problem for general nonlinear multi-agent systems is studied, where both leaderless and leader-follower cases are considered in a unified framework. The key idea is to convert consensus problems into optimal control problems where the objective of each agent with nonlinear dynamics is to design the control input minimizing the global consensus cost function. Compared with the existing distributed consensus control for nonlinear multi-agent systems, we propose a distributed optimal consensus algorithm based on the optimal control principle (OCP) method, and two enhanced algorithms are developed under the model predictive control (MPC) framework,these two algorithms demonstrate broader applicability when handling general nonlinear multi-agent systems. Moreover, the convergence and superlinear convergence rate of the proposed algorithms are rigorously analyzed. Numerical simulations demonstrate the effectiveness of the proposed algorithms.

math.OC↗

arXiv2025

IGDMRec: Behavior Conditioned Item Graph Diffusion for Multimodal Recommendation

Multimodal recommender systems (MRSs) are critical for various online platforms, offering users more accurate personalized recommendations by incorporating multimodal information of items. Structure-based MRSs have achieved state-of-the-art performance by constructing semantic item graphs, which explicitly model relationships between items based on modality feature similarity. However, such semantic item graphs are often noisy due to 1) inherent noise in multimodal information and 2) misalignment between item semantics and user-item co-occurrence relationships, which introduces false links and leads to suboptimal recommendations. To address this challenge, we propose Item Graph Diffusion for Multimodal Recommendation (IGDMRec), a novel method that leverages a diffusion model with classifier-free guidance to denoise the semantic item graph by integrating user behavioral information. Specifically, IGDMRec introduces a Behavior-conditioned Graph Diffusion (BGD) module, incorporating interaction data as conditioning information to guide the denoising of the semantic item graph. Additionally, a Conditional Denoising Network (CD-Net) is designed to implement the denoising process with manageable complexity. Finally, we propose a contrastive representation augmentation scheme that leverages both the denoised item graph and the original item graph to enhance item representations. \LL{Extensive experiments on four real-world datasets demonstrate the superiority of IGDMRec over competitive baselines, with robustness analysis validating its denoising capability and ablation studies verifying the effectiveness of its key components.

cs.IR↗

arXiv2025

An Efficient Algorithm for Learning-Based Visual Localization

This paper addresses the visual localization problem in Global Positioning System (GPS)-denied environments, where computational resources are often limited. To achieve efficient and robust performance under these constraints, we propose a novel algorithm. The algorithm stems from the optimal control principle (OCP). It incorporates diagonal information estimation of the Hessian matrix, which results in training a higher-performance deep neural network and accelerates optimization convergence. Experimental results on public datasets demonstrate that the final model achieves competitive localization accuracy and exhibits remarkable generalization capability. This study provides new insights for developing high-performance offline positioning systems.

math.OC↗

arXiv2025

Distributed Optimization Algorithm with Superlinear Convergence Rate

This paper considers distributed optimization problems, where each agent cooperatively minimizes the sum of local objective functions through the communication with its neighbors. The widely adopted distributed gradient method in solving this problem suffers from slow convergence rates, which motivates us to incorporate the second-order information of the objective functions. However, the challenge arises from the unique structure of the inverse of the Hessian matrix, which prevents the direct distributed implementation of the second-order method. We overcome this challenge by proposing a novel optimization framework. The key idea is to transform the distributed optimization problem into an optimal control problem. Using Pontryagin's maximum principle and the associated forward-backward difference equations (FBDEs), we derive a new distributed optimization algorithm that incorporates the second-order information without requiring the computation of the inverse of the Hessian matrix. Furthermore, the superlinear convergence of the proposed algorithm is proved under some mild assumptions. Finally, we also propose a variant of the algorithm to balance the number of iterations and communication.

math.OC↗

arXiv2025

SLIF-MR: Self-loop Iterative Fusion of Heterogeneous Auxiliary Information for Multimodal Recommendation

Knowledge graphs (KGs) and multimodal item information, which respectively capture relational and attribute features, play a crucial role in improving recommender system accuracy. Recent studies have attempted to integrate them via multimodal knowledge graphs (MKGs) to further enhance recommendation performance. However, existing methods typically freeze the MKG structure during training, which limits the full integration of structural information from heterogeneous graphs (e.g., KG and user-item interaction graph), and results in sub-optimal performance. To address this challenge, we propose a novel framework, termed Self-loop Iterative Fusion of Heterogeneous Auxiliary Information for Multimodal Recommendation (SLIF-MR), which leverages item representations from previous training epoch as feedback signals to dynamically optimize the heterogeneous graph structures composed of KG, multimodal item feature graph, and user-item interaction graph. Through this iterative fusion mechanism, both user and item representations are refined, thus improving the final recommendation performance. Specifically, based on the feedback item representations, SLIF-MR constructs an item-item correlation graph, then integrated into the establishment process of heterogeneous graphs as additional new structural information in a self-loop manner. Consequently, the internal structures of heterogeneous graphs are updated with the feedback item representations during training. Moreover, a semantic consistency learning strategy is proposed to align heterogeneous item representations across modalities. The experimental results show that SLIF-MR significantly outperforms existing methods, particularly in terms of accuracy and robustness.

cs.IR↗

arXiv2025

Superlinear Optimization Algorithms

This paper proposes several novel optimization algorithms for minimizing a nonlinear objective function. The algorithms are enlightened by the optimal state trajectory of an optimal control problem closely related to the minimized objective function. They are superlinear convergent when appropriate parameters are selected as required. Unlike Newton's method, all of them can be also applied in the case of a singular Hessian matrix. More importantly, by reduction, some of them avoid calculating the inverse of the Hessian matrix or an identical dimension matrix and some of them need only the diagonal elements of the Hessian matrix. In these cases, these algorithms still outperform the gradient descent method. The merits of the proposed optimization algorithm are illustrated by numerical experiments.

math.OC↗

arXiv2025

Distributed Optimization Method Based On Optimal Control

In this paper, a novel distributed optimization framework has been proposed. The key idea is to convert optimization problems into optimal control problems where the objective of each agent is to design the current control input minimizing the original objective function of itself and updated size for the future time instant. Compared with the existing distributed optimization problem for optimizing a sum of convex objective functions corresponding to multiple agents, we present a distributed optimization algorithm for multi-agents system based on the results from the maximum principle. Moreover, the convergence and superlinear convergence rate are also analyzed stringently.

math.OC↗

arXiv2024

Target-Driven Distillation: Consistency Distillation with Target Timestep Selection and Decoupled Guidance

Consistency distillation methods have demonstrated significant success in accelerating generative tasks of diffusion models. However, since previous consistency distillation methods use simple and straightforward strategies in selecting target timesteps, they usually struggle with blurs and detail losses in generated images. To address these limitations, we introduce Target-Driven Distillation (TDD), which (1) adopts a delicate selection strategy of target timesteps, increasing the training efficiency; (2) utilizes decoupled guidances during training, making TDD open to post-tuning on guidance scale during inference periods; (3) can be optionally equipped with non-equidistant sampling and x0 clipping, enabling a more flexible and accurate way for image sampling. Experiments verify that TDD achieves state-of-the-art performance in few-step generation, offering a better choice among consistency distillation models.

cs.CV↗

arXiv2023

Optimization Method Based On Optimal Control

In this paper, we focus on a method based on optimal control to address the optimization problem. The objective is to find the optimal solution that minimizes the objective function. We transform the optimization problem into optimal control by designing an appropriate cost function. Using Pontryagin's Maximum Principle and the associated forward-backward difference equations (FBDEs), we derive the iterative update gain for the optimization. The steady system state can be considered as the solution to the optimization problem. Finally, we discuss the compelling characteristics of our method and further demonstrate its high precision, low oscillation, and applicability for finding different local minima of non-convex functions through several simulation examples.

math.OC↗

arXiv2019

Grand Challenge of 106-Point Facial Landmark Localization

Facial landmark localization is a very crucial step in numerous face related applications, such as face recognition, facial pose estimation, face image synthesis, etc. However, previous competitions on facial landmark localization (i.e., the 300-W, 300-VW and Menpo challenges) aim to predict 68-point landmarks, which are incompetent to depict the structure of facial components. In order to overcome this problem, we construct a challenging dataset, named JD-landmark. Each image is manually annotated with 106-point landmarks. This dataset covers large variations on pose and expression, which brings a lot of difficulties to predict accurate landmarks. We hold a 106-point facial landmark localization competition1 on this dataset in conjunction with IEEE International Conference on Multimedia and Expo (ICME) 2019. The purpose of this competition is to discover effective and robust facial landmark localization approaches.

cs.CV↗