Institution profile

University of Science and Technology of China

Academic institutionasia · cn
Official website
Research library3,484linked papers
Opportunities0open roles
Selected work

Representative Papers

Semantic Image Synthesis via Diffusion Models

Jun 30, 2022arXiv.org

Existing GAN-based semantic image synthesis methods suffer from inherent trade-offs between generation quality and diversity. To address this, we propose the first semantic image synthesis framework built upon denoising diffusion probabilistic models (DDPMs). Our method fundamentally decouples two key inputs: noisy images are fed into the U-Net encoder, while semantic layouts guide a dedicated decoder path via multi-level Spatially-Adaptive Denormalization (SPADE). Crucially, we introduce classifier-free guidance—the first such application in semantic diffusion synthesis—to substantially improve layout-to-pixel alignment. Evaluated on four standard benchmarks—Cityscapes, ADE20K, COCO-Stuff, and Mapillary Vistas—our approach achieves state-of-the-art performance: FID of 14.3 and LPIPS of 0.52, demonstrating significant gains in both visual fidelity and semantic consistency.

170 citations26 influentialRead paper

GroupDebate: Enhancing the Efficiency of Multi-Agent Debate Using Group Discussion

Sep 21, 2024arXiv.org

To address the scalability bottleneck in multi-agent debate—specifically, the exponential growth in token consumption with increasing agent count and debate rounds—this paper proposes a *grouped multi-agent debate* architecture. Agents are partitioned into disjoint subgroups that conduct parallel internal debates; inter-group information exchange and a dynamic consensus mechanism then aggregate intermediate results efficiently. This approach breaks the traditional linear scaling constraint and represents the first systematic integration of grouping principles into multi-agent debate frameworks. Extensive experiments across multiple logical reasoning benchmarks demonstrate that our method reduces token consumption by up to 51.7% relative to baseline methods, while simultaneously improving accuracy by up to 25%. The architecture thus achieves a significant trade-off improvement between computational efficiency and reasoning performance.

37 citations2 influentialRead paper

Brightness Perceiving for Recursive Low-Light Image Enhancement

Jun 01, 2024IEEE Transactions on Artificial Intelligence

To address the severe contrast degradation and heterogeneous detail loss caused by wide dynamic range in real-world low-light scenes—challenging end-to-end enhancement methods to achieve unified improvement—this paper proposes a brightness-aware recursive enhancement framework. Methodologically, it introduces (1) a novel Brightness Perception Network (BP-Net) that dynamically determines and controls the number of recursive enhancement iterations; (2) a dual-branch Adaptive Contrast and Texture Network (ACT-Net) jointly optimizing luminance distribution and gradient-based texture fidelity; and (3) an unsupervised collaborative training strategy leveraging a self-constructed wide-luminance-distribution dataset for joint optimization. Quantitatively, the method achieves state-of-the-art performance across six reference-based and reference-free metrics, with a 0.9 dB PSNR gain. Qualitatively, it significantly improves detail preservation and visual naturalness in extremely dark and mid-to-low-light regions, demonstrating strong generalization capability.

12 citationsRead paper

TrimCaching: Parameter-sharing Edge Caching for AI Model Downloading

Apr 22, 2024arXiv.org

Existing edge caching mechanisms for AI model delivery in 5G/6G networks overlook parameter-block reuse—e.g., shared knowledge units across CNNs or LLMs—leading to low storage efficiency and limited cache hit rates under stringent latency constraints. Method: We propose a parameter-sharing-aware edge model caching framework that, for the first time, formulates parameter-block reuse as a submodular optimization problem. We design a polynomial-time algorithm with theoretical approximation guarantees and provide a general greedy solution. The framework jointly optimizes storage efficiency and service latency in multi-edge wireless networks. Results: Simulation results demonstrate that our approach significantly improves cache hit rates over conventional content-based caching, validating the effectiveness and practicality of parameter-level sharing for edge AI deployment.

12 citationsRead paper

When xURLLC Meets NOMA: A Stochastic Network Calculus Perspective

Jun 01, 2024IEEE Communications Magazine

To address the stringent requirements of ultra-low latency, ultra-high reliability, and fresh information (quantified by Age of Information, AoI) in xURLLC systems, this paper proposes a NOMA-assisted uplink architecture. It introduces stochastic network calculus (SNC) into the NOMA-xURLLC domain for the first time, establishing a unified theoretical framework that enables joint statistical QoS provisioning (SQP) for latency, AoI, and reliability tail distributions. Furthermore, we propose an SQP-driven power optimization paradigm, leveraging convex optimization and a customized power allocation algorithm to minimize uplink transmit power while satisfying multi-dimensional QoS constraints. Simulation results demonstrate that the proposed scheme outperforms conventional orthogonal multiple access across all key metrics—latency, AoI, reliability, and energy efficiency.

7 citationsRead paper
Recent publications

Latest Papers