arXiv subjects

Hao Chen

Publications and source records attributed to Hao Chen.

At least 343 records · Page 19Linked to original sources

arXiv2026

QSIM: Mitigating Overestimation in Multi-Agent Reinforcement Learning via Action Similarity Weighted Q-Learning

Value decomposition (VD) methods have achieved remarkable success in cooperative multi-agent reinforcement learning (MARL). However, their reliance on the max operator for temporal-difference (TD) target calculation leads to systematic Q-value overestimation. This issue is particularly severe in MARL due to the combinatorial explosion of the joint action space, which often results in unstable learning and suboptimal policies. To address this problem, we propose QSIM, a similarity weighted Q-learning framework that reconstructs the TD target using action similarity. Instead of using the greedy joint action directly, QSIM forms a similarity weighted expectation over a structured near-greedy joint action space. This formulation allows the target to integrate Q-values from diverse yet behaviorally related actions while assigning greater influence to those that are more similar to the greedy choice. By smoothing the target with structurally relevant alternatives, QSIM effectively mitigates overestimation and improves learning stability. Extensive experiments demonstrate that QSIM can be seamlessly integrated with various VD methods, consistently yielding superior performance and stability compared to the original algorithms. Furthermore, empirical analysis confirms that QSIM significantly mitigates the systematic value overestimation in MARL. Code is available at https://github.com/MaoMaoLYJ/pymarl-qsim.

cs.MA↗

arXiv2026

Concatenated Sum-Rank Codes

Sum-rank codes have wide applications in multishot network coding, distributed storage and the construction of space-time codes. Asymptotically good sequences of linearized algebraic geometry sum-rank codes, exceeding the Gilbert-Varshamov-like bound, were constructed in a recent paper published in IEEE Trans. Inf. Theory by E. Berardini and X. Caruso. We call this bound the Tsfasman-Vladut-Zink-like bound. In this paper, we introduce the concatenation of a sum-rank code and a Hamming metric code. Then many sum-rank codes with good parameters, which are better than sum-rank BCH codes, are constructed simply and explicitly. Moreover, we obtain an asymptotically good sequence of sum-rank codes exceeding the Tsfasman-Vladut-Zink-like bound and the Gilbert-Varshamov-like bound.

cs.IT↗

arXiv2026

Test-Time Computing for Referring Multimodal Large Language Models

We propose ControlMLLM++, a novel test-time adaptation framework that injects learnable visual prompts into frozen multimodal large language models (MLLMs) to enable fine-grained region-based visual reasoning without any model retraining or fine-tuning. Leveraging the insight that cross-modal attention maps intrinsically encode semantic correspondences between textual tokens and visual regions, ControlMLLM++ optimizes a latent visual token modifier during inference via a task-specific energy function to steer model attention towards user-specified areas. To enhance optimization stability and mitigate language prompt biases, ControlMLLM++ incorporates an improved optimization strategy (Optim++) and a prompt debiasing mechanism (PromptDebias). Supporting diverse visual prompt types including bounding boxes, masks, scribbles, and points, our method demonstrates strong out-of-domain generalization and interpretability. The code is available at https://github.com/mrwu-mac/ControlMLLM.

cs.CV↗

arXiv2026

Enhancing Large Language Models (LLMs) for Telecom using Dynamic Knowledge Graphs and Explainable Retrieval-Augmented Generation

Large language models (LLMs) have shown strong potential across a variety of tasks, but their application in the telecom field remains challenging due to domain complexity, evolving standards, and specialized terminology. Therefore, general-domain LLMs may struggle to provide accurate and reliable outputs in this context, leading to increased hallucinations and reduced utility in telecom operations.To address these limitations, this work introduces KG-RAG-a novel framework that integrates knowledge graphs (KGs) with retrieval-augmented generation (RAG) to enhance LLMs for telecom-specific tasks. In particular, the KG provides a structured representation of domain knowledge derived from telecom standards and technical documents, while RAG enables dynamic retrieval of relevant facts to ground the model's outputs. Such a combination improves factual accuracy, reduces hallucination, and ensures compliance with telecom specifications.Experimental results across benchmark datasets demonstrate that KG-RAG outperforms both LLM-only and standard RAG baselines, e.g., KG-RAG achieves an average accuracy improvement of 14.3% over RAG and 21.6% over LLM-only models. These results highlight KG-RAG's effectiveness in producing accurate, reliable, and explainable outputs in complex telecom scenarios.

cs.AI↗

arXiv2026

Strong Singleton-Like Bounds, Quasi-Perfect Codes and Distance-Optimal Codes in the Sum-Rank Metric

Codes in the sum-rank metric have received many attentions in recent years, since they have wide applications in the multishot network coding, the space-time coding and the distributed storage. In this paper, by constructing covering codes in the sum-rank metric from covering codes in the Hamming metric, we derive new upper bounds on sizes, the covering radii and the block length functions of codes in the sum-rank metric. As applications, we present several strong Singleton-like bounds that are tighter than the classical Singleton-like bound when block lengths are large. In addition, we give the explicit constructions of the distance-optimal sum-rank codes of matrix sizes $s\times s$ and $2\times 2$ with minimum sum-rank distance four respectively by using cyclic codes in the Hamming metric. More importantly, we present an infinite families of quasi-perfect $q$-ary sum-rank codes with matrix sizes $2\times m$. Furthermore, we construct almost MSRD codes with larger block lengths and demonstrate how the Plotkin sum can be used to give more distance-optimal sum-rank codes.

cs.IT↗

arXiv2026

SCENE OTA-FD: Self-Centering Noncoherent Estimator for Over-the-Air Federated Distillation

We propose SCENE (Self-Centering Noncoherent Estimator), a pilot-free and phase-invariant aggregation primitive for over-the-air federated distillation (OTA-FD). Each device maps its soft-label (class-probability) vector to nonnegative transmit energies under constant per-round power and constant-envelope signaling (PAPR near 1). At the server, a self-centering energy estimator removes the noise-energy offset and yields an unbiased estimate of the weighted soft-label average, with variance decaying on the order of 1/(SM) in the number of receive antennas M and repetition factor S. We also develop a pilot-free ratio-normalized variant that cancels unknown large-scale gains, provide a convergence bound consistent with coherent OTA-FD analyses, and present an overhead-based crossover comparison. SCENE targets short-coherence and hardware-constrained regimes, where avoiding per-round CSI is essential: it trades a modest noncoherent variance constant for zero uplink pilots, unbiased aggregation, and hardware-friendly transmission, and can outperform coherent designs when pilot overhead is non-negligible.

eess.SP↗

arXiv2026

3DGSNav: Enhancing Vision-Language Model Reasoning for Object Navigation via Active 3D Gaussian Splatting

Object navigation is a core capability of embodied intelligence, enabling an agent to locate target objects in unknown environments. Recent advances in vision-language models (VLMs) have facilitated zero-shot object navigation (ZSON). However, existing methods often rely on scene abstractions that convert environments into semantic maps or textual representations, causing high-level decision making to be constrained by the accuracy of low-level perception. In this work, we present 3DGSNav, a novel ZSON framework that embeds 3D Gaussian Splatting (3DGS) as persistent memory for VLMs to enhance spatial reasoning. Through active perception, 3DGSNav incrementally constructs a 3DGS representation of the environment, enabling trajectory-guided free-viewpoint rendering of frontier-aware first-person views. Moreover, we design structured visual prompts and integrate them with Chain-of-Thought (CoT) prompting to further improve VLM reasoning. During navigation, a real-time object detector filters potential targets, while VLM-driven active viewpoint switching performs target re-verification, ensuring efficient and reliable recognition. Extensive evaluations across multiple benchmarks and real-world experiments on a quadruped robot demonstrate that our method achieves robust and competitive performance against state-of-the-art approaches.The Project Page:https://aczheng-cai.github.io/3dgsnav.github.io/

cs.RO↗

arXiv2026

Free Lunch in Medical Image Foundation Model Pre-training via Randomized Synthesis and Disentanglement

Medical image foundation models (MIFMs) have demonstrated remarkable potential for a wide range of clinical tasks, yet their development is constrained by the scarcity, heterogeneity, and high cost of large-scale annotated datasets. Here, we propose RaSD (Randomized Synthesis and Disentanglement), a scalable framework for pre-training MIFMs entirely on synthetic data. By modeling anatomical structures and appearance variations with randomized Gaussian distributions, RaSD exposes models to sufficient multi-scale structural and appearance perturbations, forcing them to rely on invariant and task-relevant anatomical cues rather than dataset-specific textures, thereby enabling robust and transferable representation learning. We pre-trained RaSD on 1.2 million 3D volumes and 9.6 million 2D images, and extensively evaluated the resulting models across 6 imaging modalities, 48 datasets, and 56 downstream tasks. Across all evaluated downstream tasks, RaSD consistently outperforms training-from-scratch models, achieves the best performance on 17 tasks, and remains comparable to models pre-trained on large real datasets in most others. These results demonstrate that the capacity of synthetic data alone to drive robust representation learning. Our findings establish a paradigm shift in medical AI, demonstrating that synthetic data can serve as a "free lunch" for scalable, privacy-preserving, and clinically generalizable foundation models.

q-bio.QM↗

arXiv2026

GMRES Convergence Analysis for Nonsymmetric Saddle-Point Systems When the Field of Values Contains the Origin

We present a field-of-values (FOV) analysis for preconditioned nonsymmetric saddle-point linear systems, where zero is included in the field of values of the matrix. We rely on recent results of Crouzeix and Greenbaum [Spectral sets: numerical range and beyond. SIAM Journal on Matrix Analysis and Applications, 40(3):1087-1101, 2019], showing that a convex region with a circular hole is a spectral set. Sufficient conditions are derived for convergence independent of the matrix dimensions. We apply our results to preconditioned nonsymmetric saddle-point systems, and show their applicability to families of block preconditioners that have not been previously covered by existing FOV analysis. A limitation of our theory is that the preconditioned matrix is required to have a small skew-symmetric part in norm. Consequently, our analysis may not be applicable, for example, to fluid flow problems characterized by a small viscosity coefficient. Some numerical results illustrate our findings.

math.NA↗

arXiv2026

Winner Team Mia at TextVQA Challenge 2021: Vision-and-Language Representation Learning with Pre-trained Sequence-to-Sequence Model

TextVQA requires models to read and reason about text in images to answer questions about them. Specifically, models need to incorporate a new modality of text present in the images and reason over it to answer TextVQA questions. In this challenge, we use generative model T5 for TextVQA task. Based on pre-trained checkpoint T5-3B from HuggingFace repository, two other pre-training tasks including masked language modeling(MLM) and relative position prediction(RPP) are designed to better align object feature and scene text. In the stage of pre-training, encoder is dedicate to handle the fusion among multiple modalities: question text, object text labels, scene text labels, object visual features, scene visual features. After that decoder generates the text sequence step-by-step, cross entropy loss is required by default. We use a large-scale scene text dataset in pre-training and then fine-tune the T5-3B with the TextVQA dataset only.

cs.CV↗

arXiv2026

Supervised Fine-Tuning Needs to Unlock the Potential of Token Priority

The transition from fitting empirical data to achieving true human utility is fundamentally constrained by a granularity mismatch, where fine-grained autoregressive generation is often supervised by coarse or uniform signals. This position paper advocates Token Priority as the essential bridge, formalizing Supervised Fine-Tuning (SFT) not as simple optimization but as a precise distribution reshaping process that aligns raw data with the ideal alignment manifold. We analyze recent breakthroughs through this unified lens, categorizing them into two distinct regimes: Positive Priority for noise filtration and Signed Priority for toxic modes unlearning. We revisit existing progress and limitations, identify key challenges, and suggest directions for future research.

cs.CL↗

arXiv2026

SciDataCopilot: An Agentic Data Preparation Framework for AGI-driven Scientific Discovery

The current landscape of AI for Science (AI4S) is predominantly anchored in large-scale textual corpora, where generative AI systems excel at hypothesis generation, literature search, and multi-modal reasoning. However, a critical bottleneck for accelerating closed-loop scientific discovery remains the utilization of raw experimental data. Characterized by extreme heterogeneity, high specificity, and deep domain expertise requirements, raw data possess neither direct semantic alignment with linguistic representations nor structural homogeneity suitable for a unified embedding space. The disconnect prevents the emerging class of Artificial General Intelligence for Science (AGI4S) from effectively interfacing with the physical reality of experimentation. In this work, we extend the text-centric AI-Ready concept to Scientific AI-Ready data paradigm, explicitly formalizing how scientific data is specified, structured, and composed within a computational workflow. To operationalize this idea, we propose SciDataCopilot, an autonomous agentic framework designed to handle data ingestion, scientific intent parsing, and multi-modal integration in a end-to-end manner. By positioning data readiness as a core operational primitive, the framework provides a principled foundation for reusable, transferable systems, enabling the transition toward experiment-driven scientific general intelligence. Extensive evaluations across three heterogeneous scientific domains show that SciDataCopilot improves efficiency, scalability, and consistency over manual pipelines, with up to 30$\times$ speedup in data preparation.

cs.DB↗

arXiv2026

PASH at TREC 2021 Deep Learning Track: Generative Enhanced Model for Multi-stage Ranking

This paper describes the PASH participation in TREC 2021 Deep Learning Track. In the recall stage, we adopt a scheme combining sparse and dense retrieval method. In the multi-stage ranking phase, point-wise and pair-wise ranking strategies are used one after another based on model continual pre-trained on general knowledge and document-level data. Compared to TREC 2020 Deep Learning Track, we have additionally introduced the generative model T5 to further enhance the performance.

cs.IR↗

arXiv2026

Decoupled Orthogonal Dynamics: Regularization for Deep Network Optimizers

Is the standard weight decay in AdamW truly optimal? Although AdamW decouples weight decay from adaptive gradient scaling, a fundamental conflict remains: the Radial Tug-of-War. In deep learning, gradients tend to increase parameter norms to expand effective capacity while steering directions to learn features, whereas weight decay indiscriminately suppresses norm growth. This push--pull interaction induces radial oscillations, injecting noise into Adam's second-moment estimates and potentially degrading delicate tangential feature learning. We argue that magnitude and direction play distinct roles and should be decoupled in optimizer dynamics. We propose Orthogonal Dynamics Decoupling and instantiate it as AdamO: an SGD-style update handles the one-dimensional norm control, while Adam's adaptive preconditioning is confined to the tangential subspace. AdamO further incorporates curvature-adaptive radial step sizing and architecture-aware rules and projections for scale-invariant layers and low-dimensional parameters. Experiments on vision and language tasks show that AdamO improves generalization and stability over AdamW without introducing additional complex constraints.

cs.LG↗

arXiv2026

Ringmaster: How to juggle high-throughput host OS system calls from TrustZone TEEs

Many safety-critical systems require timely processing of sensor inputs to avoid potential safety hazards. Additionally, to support useful application features, such systems increasingly have a large rich operating system (OS) at the cost of potential security bugs. Thus, if a malicious party gains supervisor privileges, they could cause real-world damage by denying service to time-sensitive programs. Many past approaches to this problem completely isolate time-sensitive programs with a hypervisor; however, this prevents the programs from accessing useful OS services. We introduce Ringmaster, a novel framework that enables enclaves or TEEs (Trusted Execution Environments) to asynchronously access rich, but potentially untrusted, OS services via Linux's io_uring. When service is denied by the untrusted OS, enclaves continue to operate on Ringmaster's minimal ARM TrustZone kernel with access to small, critical device drivers. This approach balances the need for secure, time-sensitive processing with the convenience of rich OS services. Additionally, Ringmaster supports large unmodified programs as enclaves, offering lower overhead compared to existing systems. We demonstrate how Ringmaster helps us build a working highly-secure system with minimal engineering. In our experiments with an unmanned aerial vehicle, Ringmaster achieved nearly 1GiB/sec of data into enclave on a Raspberry Pi4b, 0-3% throughput overhead compared to non-enclave tasks.

cs.CR↗

arXiv2026

CAARMA: Class Augmentation with Adversarial Mixup Regularization

Speaker verification is a typical zero-shot learning task, where inference of unseen classes is performed by comparing embeddings of test instances to known examples. The models performing inference must hence naturally generate embeddings that cluster same-class instances compactly, while maintaining separation across classes. In order to learn to do so, they are typically trained on a large number of classes (speakers), often using specialized losses. However real-world speaker datasets often lack the class diversity needed to effectively learn this in a generalizable manner. We introduce CAARMA, a class augmentation framework that addresses this problem by generating synthetic classes through data mixing in the embedding space, expanding the number of training classes. To ensure the authenticity of the synthetic classes we adopt a novel adversarial refinement mechanism that minimizes categorical distinctions between synthetic and real classes. We evaluate CAARMA on multiple speaker verification tasks, as well as other representative zero-shot comparison-based speech analysis tasks and obtain consistent improvements: our framework demonstrates a significant improvement of 8\% over all baseline models. The code is available at: https://github.com/massabaali7/CAARMA/

cs.SD↗

arXiv2026

Glance and Focus Reinforcement for Pan-cancer Screening

Pan-cancer screening in large-scale CT scans remains challenging for existing AI methods, primarily due to the difficulty of localizing diverse types of tiny lesions in large CT volumes. The extreme foreground-background imbalance significantly hinders models from focusing on diseased regions, while redundant focus on healthy regions not only decreases the efficiency but also increases false positives. Inspired by radiologists' glance and focus diagnostic strategy, we introduce GF-Screen, a Glance and Focus reinforcement learning framework for pan-cancer screening. GF-Screen employs a Glance model to localize the diseased regions and a Focus model to precisely segment the lesions, where segmentation results of the Focus model are leveraged to reward the Glance model via Reinforcement Learning (RL). Specifically, the Glance model crops a group of sub-volumes from the entire CT volume and learns to select the sub-volumes with lesions for the Focus model to segment. Given that the selecting operation is non-differentiable for segmentation training, we propose to employ the segmentation results to reward the Glance model. To optimize the Glance model, we introduce a novel group relative learning paradigm, which employs group relative comparison to prioritize high-advantage predictions and discard low-advantage predictions within sub-volume groups, not only improving efficiency but also reducing false positives. In this way, for the first time, we effectively extend cutting-edge RL techniques to tackle the specific challenges in pan-cancer screening. Extensive experiments on 16 internal and 7 external datasets across 9 lesion types demonstrated the effectiveness of GF-Screen. Notably, GF-Screen leads the public validation leaderboard of MICCAI FLARE25 pan-cancer challenge, surpassing the FLARE24 champion solution by a large margin (+25.6% DSC and +28.2% NSD).

cs.CV↗

arXiv2026

Zero-Shot Knowledge Base Resizing for Rate-Adaptive Digital Semantic Communication

Digital semantic communication systems, which often leverage the Vector Quantized Variational Autoencoder (VQ-VAE) framework, are pivotal for future wireless networks. In a VQ-VAE-based semantic communication system, the transmission rate is directly governed by the size of a discrete codebook known as knowledge base (KB). However, the KB size is a fixed hyperparameter, meaning that adapting the rate requires training and storing a separate model for each desired size -- a practice that is too computationally and storage-prohibitive to achieve truly granular rate control. To address this, we introduce a principled, zero-shot KB resizing method that enables on-the-fly rate adaptation without any retraining. Our approach establishes a global importance ranking for all vectors within a single, large parent KB by uncovering its inherent semantic hierarchy. This is achieved via a three-step framework: 1) embedding KB vectors into hyperbolic space to reveal their hierarchical relationships; 2) constructing a master semantic tree using a minimum spanning tree algorithm; 3) enabling instant resizing by iteratively pruning the least important leaf nodes. Extensive simulations demonstrate that our method achieves reconstruction quality nearly identical to that of dedicated KBs trained from scratch, while demanding only a fraction of the computational budget. Moreover, our approach exhibits superior robustness at very low rates, where conventional KBs suffer from catastrophic failure. Our work resolves a fundamental limitation of VQ-VAE-based semantic communication systems, offering a practical and efficient path toward flexible and rate-adaptive semantic communication.

cs.IT↗