arXiv subjects

Yueshen Xu

Publications and source records attributed to Yueshen Xu.

14 recordsLinked to original sources

arXiv2026

ViSAGE: Constructing Self-Correcting Memories for Long-Form Video Understanding

Multimodal agents operating in long-horizon environments must build and continually update multimedia memories to support entity-consistent, temporally grounded reasoning. However, existing agentic memory approaches often discard fine-grained dentity cues under aggressive compression and segment-wise processing. They also rely heavily on vector similarity retrieval, which can surface semantically related yet identity-mismatched evidence, leading to entity confusion, error propagation, and hallucinated answers. We propose ViSAGE, a multimodal agentic memory framework that constructs self-correcting, entity-centric memories. Specifically, ViSAGE anchors entity identity via cross-modal binding over long temporal ranges. It then applies bidirectional memory refinement to propagate delayed identity evidence, retroactively unifying historical records and improving future reasoning. We also introduce multi-agent cross-verification to assess retrieved evidence under an identity-evidence alignment onstraint, enabling abstention instead of unsupported answers when evidence is missing. Extensive results demonstrate that ViSAGE consistently outperforms the strongest baseline, achieving 5.9% higher accuracy.

cs.AI↗

arXiv2026

Generation Quality-Latency Tradeoff-Aware Inference Offloading for Multimodal LLMs in Cloud-Edge Continuum

Beyond pure cloud, some efforts are being made to deploy Large Language Models (LLMs) in edge to accelerate inference response. So the deployment of LLMs in cloud-edge continuum becomes a promising paradigm, where the tasks involving multimodal data occupy a large part of requests. Under this continuum, users usually concern about multiple Quality-of-Service (QoS) attributes, but it is always intractable to jointly optimize them. In this paper, we propose to study the joint optimization of those attributes and focus on two key representatives, i.e., content generation quality and response latency. We propose to study the offloading technology to achieve a tradeoff between the two objectives in the cloud-edge collaborative Multimodal LLM (MLLM) system. However, it is highly difficult to predict generation quality and inference latency for MLLM inference tasks while optimizing this offloading process. To address these unprecedented difficulties, we propose a Quality-Latency Tradeoff-Aware MLLM Inference Offloading (QLMIO) framework to make decisions that optimally balance generation quality and response latency. Meanwhile, recognizing the absence of publicly available datasets tailored to the MLLM inference offloading problem, we constructed a real-world cloud-edge collaborative MLLM system and subsequently collected an MLLM Inference Offloading Benchmark (MIOBench) to comprehensively evaluate our framework and facilitate the study of this problem. Extensive experimental results demonstrate that the QLMIO framework reduces latency by up to 58.14\% compared to baselines, while simultaneously matching the task completion rate achieved under the case that executes all requests exclusively on a cloud server. The dataset and codes are available at Github.

cs.DC↗

arXiv2026

Fairness-Aware and Latency-Controllable Scheduling for Chunked-Prefill LLM Serving

As large language models (LLMs) are increasingly deployed with highly heterogeneous workloads, chunked-prefill execution has emerged as a mainstream serving architecture. Balancing scheduling fairness and latency stability in such environments is critical; otherwise, severe head-of-line blocking and request starvation will degrade user experience. However, existing systems rely on rigid First-Come, First-Served (FCFS) policies and static token budgets, leading to fairness degradation and unpredictable latency jitter. To address these issues, we propose a fairness-aware and latency-controllable scheduling framework for chunked-prefill LLM engines. Specifically, we design a lightweight aging-based scheduling policy that dynamically calculates priorities using accumulated waiting time and remaining prefill work. Furthermore, we develop Latency-Prediction-Based Request Scheduling (LPRS) and Active Prefill Control (APC) to replace static budgets with target-time constraints and actively regulate prefill concurrency. We evaluated our scheduling framework on NVIDIA GPUs and Ascend accelerators using real-world workloads. Results show the aging policy reduces mean end-to-end latency by over 10\% compared to FCFS. Moreover, LPRS and APC significantly reduce P99 tail latency and suppress prefill fragmentation, confirming that the structural prefill control and the temporal latency constraints are fundamentally complementary. All codes have been released in Github.

cs.DC↗

arXiv2026

ATOM: Instantiating Budget-Controllable Multi-Agent Collaboration via Nucleus-Electron Hierarchy

Large Language Model (LLM)-based multi-agent systems rely on optimized collaboration topologies to balance performance and communication costs. However, current methods struggle with the inherent stability-extensibility trade-off and often misalign computational budgets with query difficulty. We propose \textsc{ATOM}, an adaptive framework that generates budget-controllable collaboration graphs via a novel task-driven reinforcement learning paradigm. Inspired by atomic structures, \textsc{ATOM} employs a nucleus-electron hierarchy: it maintains a stable, offline-learned collaboration backbone (the nucleus) while dynamically activating query-conditioned agents (electrons) during inference. Crucially, a complexity-aware budgeting strategy aligns resource consumption with task demands by estimating query difficulty to strictly regulate electron instantiation. Extensive experiments across six diverse benchmarks demonstrate that \textsc{ATOM} achieves state-of-the-art performance while improving token efficiency by up to $30\%$ compared to strong baselines.

cs.MA↗

arXiv2025

Video-QTR: Query-Driven Temporal Reasoning Framework for Lightweight Video Understanding

The rapid development of multimodal large-language models (MLLMs) has significantly expanded the scope of visual language reasoning, enabling unified systems to interpret and describe complex visual content. However, applying these models to long-video understanding remains computationally intensive. Dense frame encoding generates excessive visual tokens, leading to high memory consumption, redundant computation, and limited scalability in real-world applications. This inefficiency highlights a key limitation of the traditional process-then-reason paradigm, which analyzes visual streams exhaustively before semantic reasoning. To address this challenge, we introduce Video-QTR (Query-Driven Temporal Reasoning), a lightweight framework that redefines video comprehension as a query-guided reasoning process. Instead of encoding every frame, Video-QTR dynamically allocates perceptual resources based on the semantic intent of the query, creating an adaptive feedback loop between reasoning and perception. Extensive experiments across five benchmarks: MSVD-QA, Activity Net-QA, Movie Chat, and Video MME demonstrate that Video-QTR achieves state-of-the-art performance while reducing input frame consumption by up to 73%. These results confirm that query-driven temporal reasoning provides an efficient and scalable solution for video understanding.

cs.CV↗

arXiv2025

Completion by Comprehension: Guiding Code Generation with Multi-Granularity Understanding

As code completion task from function-level to repository-level, leveraging contextual information from large-scale codebases becomes a core challenge. However, existing retrieval-augmented generation (RAG) methods typically treat code as plain natural language, relying primarily on shallow semantic matching while overlooking structural semantics and code-specific dependencies. This limits their ability to capture control flow and underlying intent, ultimately constraining the quality of generated code. Therefore, we propose CoCo, a novel framework that enables code Completion by Comprehension of multi-granularity context from large-scale code repositories. CoCo employs static code analysis to extract structured context at the function, file, and project levels, capturing execution logic and semantic dependencies. It then adopts an graph-based multi-granularity context selection mechanism to filter out redundant information and remove noise. Consequently, the information is converted into natural language in a consistent manner, thereby functioning as explicit contextual prompts to guide subsequent code completion. Additionally, a structure-aware code re-ranker mechanism ensures alignment at both semantic and structural levels. Extensive experiments on CrossCodeEval and RepoEval benchmarks demonstrate that CoCo consistently surpasses state-of-the-art baselines, achieving up to 20.2% gains in EM. Moreover, the framework is model-agnostic and can be seamlessly integrated into existing methods, leading to significant performance.

cs.SE↗

arXiv2025

MUSE: A Heterogeneity-Aware Multimedia Search Engine for Mobile SoCs

On-device multimedia retrieval is vital for smartphones, enabling applications like cross-modal semantic search and multimodal personal AI agents. However, realizing efficient retrieval on mobile SoCs remains a critical systems challenge. Unlike text-based search, multimedia applications depend on high-dimensional cross-modal embeddings (often exceeding 1024 dimensions) and continuously expanding media streams (e.g., video lifelogs and ambient audio). Mobile devices must process these intensive workloads under strict latency, energy, and memory constraints. Existing vector retrieval systems, designed primarily for static server datasets, are poorly suited for the dynamic execution models of mobile SoCs, where interactive queries must coexist with continuous background media ingestion and index maintenance. To address these challenges, we propose MUSE, a MUltimedia Search Engine comprehensively co-designed for mobile SoCs. MUSE tackles these issues through an architecture-index co-design. It introduces a three-stage asynchronous execution pipeline that overlaps DMA transfers, NPU matrix computation, and hardware-vector data adaptation to overcome the memory wall of high-dimensional embeddings. Furthermore, it redesigns the Inverted File Index (IVF) and introduces workload-aware heterogeneous scheduling across the CPU, GPU, and NPU for hybrid interactive-and-ingestion workloads. Evaluated on Snapdragon 8-series SoCs using real-world multimodal datasets, MUSE improves query throughput by up to 1.4x at matched recall, achieves up to 7x faster index construction, and delivers up to 6x higher insertion throughput under concurrent streaming. Crucially, its accelerator-native design ensures strict physical compliance, capping peak device temperatures at 38{\deg}C and reducing total energy consumption by up to 4.6x compared to CPU-bound baselines.

cs.DC↗

arXiv2025

DRAMA: Next-Gen Dynamic Orchestration for Resilient Multi-Agent Ecosystems in Flux

Multi-agent systems (MAS) have demonstrated significant effectiveness in addressing complex problems through coordinated collaboration among heterogeneous agents. However, real-world environments and task specifications are inherently dynamic, characterized by frequent changes, uncertainty, and variability. Despite this, most existing MAS frameworks rely on static architectures with fixed agent capabilities and rigid task allocation strategies, which greatly limits their adaptability to evolving conditions. This inflexibility poses substantial challenges for sustaining robust and efficient multi-agent cooperation in dynamic and unpredictable scenarios. To address these limitations, we propose DRAMA: a Dynamic and Robust Allocation-based Multi-Agent System designed to facilitate resilient collaboration in rapidly changing environments. DRAMA features a modular architecture with a clear separation between the control plane and the worker plane. Both agents and tasks are abstracted as resource objects with well-defined lifecycles, while task allocation is achieved via an affinity-based, loosely coupled mechanism. The control plane enables real-time monitoring and centralized planning, allowing flexible and efficient task reassignment as agents join, depart, or become unavailable, thereby ensuring continuous and robust task execution. The worker plane comprises a cluster of autonomous agents, each with local reasoning, task execution, the ability to collaborate, and the capability to take over unfinished tasks from other agents when needed.

cs.MA↗

arXiv2025

TRAIL: Joint Inference and Refinement of Knowledge Graphs with Large Language Models

Recent advances in large language models (LLMs) have unlocked powerful reasoning and decision-making capabilities. However, their inherent dependence on static parametric memory fundamentally limits their adaptability, factual accuracy, and interpretability in knowledge-intensive scenarios. Knowledge graphs (KGs), as structured repositories of explicit relational knowledge, offer a promising approach for augmenting LLMs with external, interpretable memory. Nevertheless, most existing methods that combine LLMs with KGs treat reasoning and knowledge updating as separate processes, resulting in suboptimal utilization of new information and hindering real-time updates. In this work, we propose TRAIL: a novel, unified framework for Thinking, Reasoning, And Incremental Learning that couples joint inference and dynamic KG refinement with large language models. TRAIL enables LLM agents to iteratively explore, update, and refine knowledge graphs during the reasoning process, employing a confidence-driven mechanism for the generation, validation, and pruning of new facts. This plug-and-play architecture facilitates seamless integration with various LLMs, supporting continual adaptation without the need for retraining. Extensive experiments on multiple benchmarks demonstrate that TRAIL outperforms existing KG-augmented and retrieval-augmented LLM baselines by 3% to 13%. More importantly, these results represent a significant step toward developing adaptive, memory-augmented language models capable of continual learning and reliable, transparent reasoning.

cs.IR↗

arXiv2025

StackPilot: Autonomous Function Agents for Scalable and Environment-Free Code Execution

Recent advances in large language models (LLMs) have substantially enhanced automated code generation across a wide range of programming languages. Nonetheless, verifying the correctness and executability of LLM-generated code remains a significant challenge, as traditional methods rely on language-specific compilers and environment-dependent runtimes. To overcome these limitations, we introduce StackPilot, an LLM-native, multi-agent framework designed for language-agnostic code verification and execution, which operates independently of conventional toolchains. StackPilot offers three principal innovations: (1) a Function-as-Agents paradigm, in which each function is modeled as an autonomous agent capable of fine-grained reasoning and collaborative verification; (2) an LLM-as-Executor strategy, which enables scalable verification via stack-based scheduling; and (3) a novel snapshot mechanism that preserves complete execution contexts, facilitating deterministic and lossless context switching during verification. Empirical evaluations demonstrate that StackPilot achieves framework reliability rates between 89% and 97%, substantially outperforming baseline approaches. These results indicate that StackPilot can reliably verify and execute a significantly larger proportion of LLM-generated code across diverse programming tasks compared to existing methods.

cs.PL↗

arXiv2025

LightRouter: Towards Efficient LLM Collaboration with Minimal Overhead

The rapid advancement of large language models has unlocked remarkable capabilities across a diverse array of natural language processing tasks. However, the considerable differences among available LLMs-in terms of cost, performance, and computational demands-pose significant challenges for users aiming to identify the most suitable model for specific tasks. In this work, we present LightRouter, a novel framework designed to systematically select and integrate a small subset of LLMs from a larger pool, with the objective of jointly optimizing both task performance and cost efficiency. LightRouter leverages an adaptive selection mechanism to identify models that require only a minimal number of boot tokens, thereby reducing costs, and further employs an effective integration strategy to combine their outputs. Extensive experiments across multiple benchmarks demonstrate that LightRouter matches or outperforms widely-used ensemble baselines, achieving up to a 25% improvement in accuracy. Compared with leading high-performing models, LightRouter achieves comparable performance while reducing inference costs by up to 27%. Importantly, our framework operates without any prior knowledge of individual models and relies exclusively on inexpensive, lightweight models. This work introduces a practical approach for efficient LLM selection and provides valuable insights into optimal strategies for model combination.

cs.AI↗

arXiv2021

Defeating Catastrophic Forgetting via Enhanced Orthogonal Weights Modification

The ability of neural networks (NNs) to learn and remember multiple tasks sequentially is facing tough challenges in achieving general artificial intelligence due to their catastrophic forgetting (CF) issues. Fortunately, the latest OWM Orthogonal Weights Modification) and other several continual learning (CL) methods suggest some promising ways to overcome the CF issue. However, none of existing CL methods explores the following three crucial questions for effectively overcoming the CF issue: that is, what knowledge does it contribute to the effective weights modification of the NN during its sequential tasks learning? When the data distribution of a new learning task changes corresponding to the previous learned tasks, should a uniform/specific weight modification strategy be adopted or not? what is the upper bound of the learningable tasks sequentially for a given CL method? ect. To achieve this, in this paper, we first reveals the fact that of the weight gradient of a new learning task is determined by both the input space of the new task and the weight space of the previous learned tasks sequentially. On this observation and the recursive least square optimal method, we propose a new efficient and effective continual learning method EOWM via enhanced OWM. And we have theoretically and definitively given the upper bound of the learningable tasks sequentially of our EOWM. Extensive experiments conducted on the benchmarks demonstrate that our EOWM is effectiveness and outperform all of the state-of-the-art CL baselines.

cs.LG↗

arXiv2021

Eliminating the Barriers: Demystifying Wi-Fi Baseband Design and Introducing the PicoScenes Wi-Fi Sensing Platform

The research on Wi-Fi sensing has been thriving over the past decade but the process has not been smooth. Three barriers always hamper the research: unknown baseband design and its influence, inadequate hardware, and the lack of versatile and flexible measurement software. This paper tries to eliminate these barriers through the following work. First, we present an in-depth study of the baseband design of the Qualcomm Atheros AR9300 (QCA9300) NIC. We identify a missing item of the existing CSI model, namely, the CSI distortion, and identify the baseband filter as its origin. We also propose a distortion removal method. Second, we reintroduce both the QCA9300 and software-defined radio (SDR) as powerful hardware for research. For the QCA9300, we unlock the arbitrary tuning of both the carrier frequency and bandwidth. For SDR, we develop a high?performance software implementation of the 802.11a/g/n/ac/ax baseband, allowing users to fully control the baseband and access the complete physical-layer information. Third, we release the PicoScenes software, which supports concurrent CSI measure?ment from multiple QCA9300, Intel Wireless Link (IWL5300) and SDR hardware. PicoScenes features rich low-level controls, packet injection and software baseband implementation. It also allows users to develop their own measurement plugins. Finally, we report state-of-the-art results in the extensive evaluations of the PicoScenes system, such as the >2 GHz available spectrum on the QCA9300, concurrent CSI measurement, and up to 40 kHz and 1 kHz CSI measurement rates achieved by the QCA9300 and SDR. PicoScenes is available at https://ps.zpj.io.

cs.AR↗

arXiv2020

COVID-19 Evolves in Human Hosts

Today, we are all threatened by an unprecedented pandemic: COVID-19. How different is it from other coronaviruses? Will it be attenuated or become more virulent? Which animals may be its original host? In this study, we collected and analyzed nearly thirty thousand publicly available complete genome sequences for COVID-19 virus from 79 different countries, the previously known flu-causing coronaviruses (HCov-229E, HCov-OC43, HCov-NL63 and HCov-HKU1) and the lethal, pathogenic viruses, SARS, MERS, Victoria, Lassa, Yamagata, Ebola, and Dengue. We found strong similarities between the current circulating COVID-19 and SARS and MERS, as well as COVID-19 in rhinolophines and pangolins. On the contrary, COVID-19 shares little similarity with the flu-causing coronaviruses and the other known viruses. Strikingly, we observed that the divergence of COVID-19 strains isolated from human hosts has steadily increased from December 2019 to May 2020, suggesting COVID-19 is actively evolving in human hosts. In this paper, we first propose a novel MLCS algorithm NP-MLCS1 for the big sequence analysis, which can calculate the common model for COVID-19 complete genome sequences to provide important information for vaccine and antibody development. Geographic and time-course analysis of the evolution trees of the human COVID-19 reveals possible evolutional paths among strains from 79 countries. This finding has important implications to the management of COVID-19 and the development of vaccines and medications.

q-bio.PE↗