arXiv subjects

Zijian Zhang

Publications and source records attributed to Zijian Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

$α$-Wasserstein Mechanism for Rényi Pufferfish Privacy

This paper introduces the $α$-Wasserstein mechanism for achieving Rényi Pufferfish Privacy using Laplace and Gaussian noise. By leveraging Hölder's inequality, we demonstrate that the scale parameter of the Laplace mechanism can be calibrated via an upper bound on the $W_α$ metric to satisfy $(α, ε)$-Rényi Pufferfish Privacy for $α\in (1, \infty]$. We show that at the limit $α= \infty$, this framework recovers the established $W_\infty$ mechanism for $ε$-pufferfish privacy. This result is subsequently extended to the exponential mechanism. Furthermore, we propose a $W_α$ mechanism for Gaussian noise for $α\in (1, \infty)$, demonstrating that it generalizes existing results within the Rényi Differential Privacy framework. Experimental evaluations reveal that our $α$-Wasserstein mechanism significantly reduces noise power compared to the conventional $W_\infty$-based approach, with the Gaussian mechanism providing superior utility over the Laplace mechanism. Notably, the mechanisms derived in this work achieve exact $(α, ε)$-Rényi Pufferfish Privacy without requiring additional relaxations, such as $δ$-approximations.

cs.CR↗

arXiv2026

X-MoD: Practical Scaling Laws for Sparse-Depth Routing Beyond Mixture-of-Depths

Mixture-of-Depths (MoD) enables conditional computation across Transformer depth by routing only a subset of tokens through selected layers, but its original one-sparse--one-dense alternation tightly couples total capacity to active capacity and limits sparse-depth scaling. We introduce X-MoD, a scalable sparse-depth architecture that decouples token sparsity from anchor stride, allowing total parameter count to grow while keeping active-equivalent capacity nearly fixed. To make deep sparse routing trainable, X-MoD combines dense anchors with variance-scaled layer-wise gating and depth-wise token balancing. To make this regime analyzable and usable, we formulate sparse-depth routing as a conditional architecture-design problem: given compute, context length, and active-equivalent backbone size, how should the routing configuration be chosen? We develop a practical scaling-law framework by fitting X-MoD relative to FLOP-matched dense baselines, yielding an interpretable law that decomposes performance into sparse-capacity gain, sparse-context correction, and anchor-stride interaction. The law predicts validation loss across routing configurations and reveals how context length, model scale, and anchor stride shape sparse-depth performance. We validate the architecture and law through pretraining sweeps, held-out scaling-law prediction, ablations, downstream evaluations, and comparisons with Dense, MoD, and representative MoE baselines.

cs.LG↗

arXiv2026

Concurrent Coded Signal-Multiplexing Ranging for Half-Duplex Asynchronous Networks

Signal-multiplexing network ranging (SM-NR) shares broadcasts across node pairs, but its sequential operation leads to a ranging cycle that grows linearly with network size. This paper proposes a concurrent coded SM-NR (CC-SM-NR) framework for asynchronous half-duplex networks. Firstly, the CC-SM-NR protocol coordinates concurrent transmissions through binary transmit-listen codewords. The transmit-listen schedule defined by these codewords ensures reciprocal observations subject to a finite concurrency limit. Then, we derive the exact minimum number of transmit-listen rounds without a concurrency limit, which reveals that the minimum grows logarithmically with network size. To account for practical scenarios, we establish the necessary and sufficient conditions for the constant-weight feasibility of codewords under a finite concurrency limit. Subsequently, we propose a low-complexity scheduling algorithm that achieves the minimum round count within the constant-weight codeword class. To support higher observation redundancy, this scheduling design is extended through a greedy construction. Finally, simulation results demonstrate the effectiveness of the proposed schemes for network ranging.

cs.IT↗

arXiv2026

ChartJudgeBench: Evaluating LMM Judges for Chart-to-Code Generation

Building strong chart-to-code systems increasingly relies on reinforcement learning, whose effectiveness depends critically on the quality of the reward signal. Large Multimodal Models (LMMs) play a natural critical role in jointly assessing chart visual appearance and task requirements. They are therefore increasingly used as visual critics and reward models, yet their reliability as judges remains largely unexplored. To this end, we introduce ChartJudgeBench, a diagnostic vision-language benchmark for assessing LMM judges in chart-to-code workflows. It includes 1,003 Chart Perception Alignment (CPA) instances for pairwise chart comparison and 650 Chart Reasoning Judgment (CRJ) instances for binary Accept/Reject verification in Chart Reproduction and Chart Editing. Together, these tasks emulate the core judging decisions required in agentic refinement and RL-based chart optimization. Our evaluation of strong LMMs reveals four systematic limitations: (i) positional bias in pairwise comparison, (ii) a strong tendency to overpredict Accept, (iii) difficulty in matching visual styles and aesthetics, and (iv) an unexpected leniency bias in RL-trained models. These findings show that current LMM judges require explicit reliability validation before being used as critics or reward models in chart-to-code optimization. The code and data are available on ChartJudgeBench.

cs.CV↗

arXiv2026

LATED: Ly$α$-anchored photometric selection of candidate metal-free and extremely metal-poor star formation from the end of reionisation to cosmic noon

Cosmological simulations allow a low-level tail of Population III (Pop III) star formation to persist to $z=2-6$. Spectroscopic confirmation is expensive, so an efficient photometric pre-selection is needed. We present LATED (Lyman-Alpha Tomography of Extremely Metal-poor Domains), which selects metal-free and extremely metal-poor candidates from a Ly$α$-emitter parent sample using strong-line diagnostics. The method requires three bands and two colours, $x=m_{\rm OIII}-m_{{\rm H}α}$ and $y=m_{{\rm H}α}-m_{\rm cont}$, which trace oxygen abundance and the H$α$ equivalent width, respectively. Requiring that the filters simultaneously contain [O III]+H$β$ and H$α$, together with a Ly$α$ parent selection, defines five windows spanning $z=1.92-6.60$ (four JWST/NIRCam, one Roman/WFI). The criteria, $x\geq x_{\rm min}(z)$ and $y\leq y_{\rm max}(z)$, are set by the per-redshift extrema of a forward-modelled Pop III template locus. Ordinary metal-enriched star-forming and AGN templates fall outside the selection region, and possible contaminants such as little red dots are flagged by their multi-band colours. To check contamination empirically, we apply LATED to 1126 JADES spectroscopic galaxies, and no source is selected in the four NIRCam windows. We release a Python package which converts the same photometry into R3=[O III]/H$β$ as a measurement or upper limit, reproducing JADES spectroscopy with small 0.12 dex scatter. Applied to 85 archival MUSE Ly$α$ emitters in Abell 2744, LATED recovers the confirmed extremely metal-poor galaxy AMORE6 and reveals three new candidates at $z=3-5$. Photometry alone cannot establish a metal-free nature. LATED delivers prioritised candidates for spectroscopic follow-up, providing a scalable route toward a systematic census of late-time Pop III star formation.

astro-ph.GA↗

arXiv2026

LATED: JWST integral field spectroscopy of a galaxy caught in chemical infancy at $z=4.8$ behind Abell 2744

When Population III (Pop III) star formation ended remains an open question. LATED-1 is an intrinsically faint ($M_\mathrm{UV}=-16.15$) Ly$α$ emitter at $z=4.80$ revealed by VLT/MUSE behind the lensing cluster Abell 2744 ($z=0.308$). Before any spectroscopic metallicity constraints, it was identified as an extremely metal-poor or metal-free galaxy candidate from JWST imaging by LATED, our novel photometric selection framework. Here we present serendipitous JWST/NIRSpec PRISM integral-field spectroscopy of this target. The spectrum reveals Ly$α$, H$β$, and H$α$ at 7.0, 5.6, and 17.7$σ$, as well as tentative detections of [O III]$λ\lambda4959,5007$ at 2.9$σ$, and yields $R3=\mathrm{[O\,III]\lambda5007/Hβ}=0.59^{+0.27}_{-0.21}$, upholding the earlier LATED photometric prediction of $R3<1.54$ ($2σ$ limit). The canonical JWST-based strong-line calibration, extrapolated to low metallicity, implies $\log (O/H)=6.45^{+0.17}_{-0.19}$, or $Z/Z_\odot=0.58_{-0.20}^{+0.28}\%$, placing LATED-1 among the most metal-poor galaxies known. Its modest magnification ($μ=2.80$) leaves the intrinsic properties insensitive to the lens model. LATED-1 is therefore a dwarf galaxy caught in the earliest state of chemical enrichment, and a compelling target for testing whether Pop III star formation can persist to $z<5$. The result demonstrates that photometric selection, especially through the LATED methodology, can reach below one per cent solar metallicity to identify Pop III galaxy candidates for spectroscopic follow-up.

astro-ph.GA↗

arXiv2026

Towards Fully Automated Medical Imaging Code Generation via Validation-based Context Engineering

Large language models (LLMs) have demonstrated considerable promise in program generation for small-scale and conventional application development; however, they remain limited when applied to complex, domain-specific tasks such as medical image processing. General-purpose models lack explicit domain knowledge and robust validation mechanisms to ensure correctness, often requiring substantial human intervention to produce reliable processing pipelines. To address these limitations, we propose AutoMedImg, a multi-agent framework for fully automated medical image processing code generation. AutoMedImg orchestrates specialised agents across two phases: a Planning Phase that performs dataset analysis and architecture design with semantic and formal verification, and a Coding Phase that generates modules in parallel with static checking, execution testing, and assembly validation. This multi-stage validation mitigates error propagation throughout generation, while comprehensive auto-context engineering combining domain-specific knowledge bases, shared memory, and validation feedback automates context construction without manual prompting. A cross-project adaptive pipeline synthesis mechanism further accumulates validated pipelines and retrieves proven components for new tasks based on project similarity, enhancing generation efficiency through cross-project learning. Extensive evaluation across six diverse and well-established medical imaging datasets with five backbone LLMs demonstrates that AutoMedImg achieves zero human intervention, with Dice scores of up to 0.90 for segmentation tasks and 99% accuracy for classification.

cs.CV↗

arXiv2026

GaussianDream++: Efficient 3D Gaussian World Modeling for Robotic Manipulation

Vision-Language-Action (VLA) policies have advanced language-conditioned robotic manipulation, yet action-imitation objectives provide only weak supervision for metric 3D structure and short-horizon physical evolution. Geometry-enhanced policies mainly improve current-scene grounding, whereas predictive policies often model future dynamics in RGB or latent spaces and may incur substantial deployment cost. GaussianDream demonstrates that training-time current Gaussian reconstruction and future Gaussian prediction provide effective 3D supervision, but its dense VGGT/TGE-based prefix jointly carries state, dynamics, and action-conditioning information. We present \textbf{\methodname}, a compact, policy-native extension that inserts \textbf{World State Tokens} and \textbf{World Prediction Tokens} directly into the VLA backbone. A training-only \textbf{World Representation Head} decodes these tokens into a Current World and coupled Future Prediction over shared Gaussian primitives, while static--dynamic factorization preserves persistent structure and focuses residual motion on interaction-relevant regions. At inference, the head, renderer, auxiliary objectives, and VGGT/TGE pathway are removed, leaving only 20 world tokens without online Gaussian decoding or rollout. \method achieves \textbf{98.6\%} on LIBERO and \textbf{87.8\%} on LIBERO-Plus, with clear gains under Camera and Layout shifts. Real-robot experiments further improve average success from 29.2\% to 52.5\% over reproduced $π_{0.5}$ while maintaining efficient closed-loop control.

cs.RO↗

arXiv2026

GaussianWAM: Distilling Geometry and Semantics from 3D Gaussian Fields into World-Action Models

World-Action Models (WAMs) jointly learn future visual prediction and action generation, using video dynamics as a representation-learning signal for robotic manipulation. However, their video latents are primarily optimized for visual prediction and are not explicitly encouraged to preserve cross-view geometric structure or spatially localized, object-relevant semantics. We propose \textbf{GaussianWAM}, a training-time representation-enhancement framework that organizes geometric and semantic supervision through a 3D Gaussian field. Given synchronized multi-view observations, frozen geometry and vision foundation models provide depth, camera parameters, and dense semantic features. GaussianWAM binds these heterogeneous signals to shared Gaussian primitives and renders spatially aligned semantic, depth, and coverage targets, which are distilled into the current-observation representations of the WAM. All teacher models, Gaussian components, and auxiliary prediction heads are removed after training, leaving the original WAM inference path without additional modules or forward computation. On LIBERO-Plus, GaussianWAM improves FastWAM from 52.05\% to 71.29\% and Cosmos Policy from 71.52\% to 77.30\%. Direct CLIP and VGGT distillation already establishes a strong FastWAM baseline of 69.37\%, while Gaussian-field unification further improves it to 71.29\%, supporting the benefit of spatially organizing heterogeneous teacher signals. GaussianWAM also improves performance on standard LIBERO and shows positive transfer trends on RoboTwin and real-world manipulation. These results suggest that training-time Gaussian distillation provides a practical way to inject geometry- and semantics-related supervision into WAM representations without changing their deployment architecture.

cs.RO↗

arXiv2026

ProtoFlow: Mitigating Forgetting in Class-Incremental Remote Sensing Segmentation via Low-Curvature Prototype Flow

Remote sensing segmentation in real deployment is inherently continual: new semantic categories emerge, and acquisition conditions shift across seasons, cities, and sensors. Despite recent progress, many incremental approaches still treat training steps as isolated updates, which leaves representation drift and forgetting insufficiently controlled. We present ProtoFlow, a time-aware prototype dynamics framework that models class prototypes as trajectories and learns their evolution with an explicit temporal vector field. By jointly enforcing low-curvature motion and inter-class separation, ProtoFlow stabilizes prototype geometry throughout incremental learning. Experiments on standard class- and domain-incremental remote sensing benchmarks show consistent gains over strong baselines, including up to 1.5-2.0 points improvement in mIoUall, together with reduced forgetting. These results suggest that explicitly modeling temporal prototype evolution is a practical and interpretable strategy for robust continual remote sensing segmentation. Open-source code:https://github.com/dudududke/protoflow.

cs.CV↗

arXiv2026

A Dark Matter Masquerade: Degeneracies in Black Hole and Accretion Inference from X-ray Reflection Measurements and Prospects for Compact Dark Matter Halo Constraints

X-ray reflection spectroscopy is an established probe of black hole (BH) spin and accretion geometry, and many studies have examined its systematic uncertainties. In this work, we explore a possible environmental effect associated with compact dark matter (DM) halos. We adopt numerical BH-DM spacetimes and perform general-relativistic ray-tracing calculations to generate broadened Fe\, K$α$ line profiles. For the configurations studied here, compact DM halos shift the lines toward lower observed energies relative to their Kerr counterparts. We fit the simulated profiles with standard Kerr models and find that the inferred spin and inclination can differ from their input values; in some cases, the spin is overestimated. These results suggest that compact DM halos may introduce an additional uncertainty in reflection-based measurements. Conversely, reflection models that include the environment could ultimately help constrain key halo parameters, such as its mass and characteristic scale, complementing gravitational-wave probes.

astro-ph.HE↗

arXiv2026

StitchCUDA: An Automated Multi-Agents End-to-End GPU Programing Framework with Rubric-based Agentic Reinforcement Learning

Modern machine learning (ML) workloads increasingly rely on GPUs, yet achieving high end-to-end performance remains challenging due to dependencies on both GPU kernel efficiency and host-side settings. Although LLM-based methods show promise on automated GPU kernel generation, prior works mainly focus on single-kernel optimization and do not extend to end-to-end programs, hindering practical deployment. To address the challenge, in this work, we propose StitchCUDA, a multi-agent framework for end-to-end GPU program generation, with three specialized agents: a Planner to orchestrate whole system design, a Coder dedicated to implementing it step-by-step, and a Verifier for correctness check and performance profiling using Nsys/NCU. To fundamentally improve the Coder's ability in end-to-end GPU programming, StitchCUDA integrates rubric-based agentic reinforcement learning over two atomic skills, task-to-code generation and feedback-driven code optimization, with combined rubric reward and rule-based reward from real executions. Therefore, the Coder learns how to implement advanced CUDA programming techniques (e.g., custom kernel fusion, cublas epilogue), and we also effectively prevent Coder's reward hacking (e.g., just copy PyTorch code or hardcoding output) during benchmarking. Experiments on KernelBench show that StitchCUDA achieves nearly 100% success rate on end-to-end GPU programming tasks, with 1.72x better speedup over the multi-agent baseline and 2.73x than the RL model baselines. Code of the STITCHCUDA framework is avalaible at https://github.com/UMN-APEX-Lab/StitchCUDA.

cs.MA↗

arXiv2026

MotionCraft: Latent World Modeling with Sparse Attention for Visual Upscaling

Video super-resolution (VSR) aims to recover high-fidelity high-resolution videos from low-resolution inputs and is central to applications ranging from mobile capture to streaming and archival restoration. Existing approaches trade off among local-detail fidelity, long-range spatio-temporal modeling, perceptual realism, and efficiency: convolutional alignment techniques preserve local structure but suffer when motion is large or degradations are complex; transformer-based methods capture long-range dependencies yet require architectural or algorithmic adaptations to remain computationally feasible; and recent latent or diffusion-based generators synthesize rich texture but require specialized temporal constraints to maintain coherence. We present MotionCraft, a controllable VSR framework that formulates restoration as motion-aware latent state prediction inspired by world models and integrates adaptive sparse attention with an explicit user-accessible control interface. MotionCraft combines robust motion fusion, a Latent World Transformer that balances locality and targeted non-local interactions, and a compact conditional decoder to deliver temporally consistent, high-quality reconstructions under streaming constraints. Empirical evaluations show that MotionCraft achieves strong reconstruction and perceptual performance while enabling predictable trade-offs between temporal smoothness and reconstruction fidelity.

cs.CV↗

arXiv2026

FreSH: Frequency-Segmented Hierarchical Multi-Expert Framework for Multivariate Time Series Classification

Multivariate Time Series Classification (MTSC) demands models that can effectively capture complex temporal patterns across multiple scales while remaining computationally efficient. However, existing approaches generally struggle to reconcile fine-grained representation learning, especially under class imbalance and real-world constraints. In this paper, we present FreSH, a Frequency-Segmented Hierarchical Multi-Expert Framework designed to address these challenges. FreSH introduces a new perspective for MTSC by enabling adaptive, multi-scale analysis of temporal signals, allowing different aspects of the data to be modeled in a complementary and coordinated manner. By combining localized specialization with holistic context modeling, FreSH achieves strong representational capacity without incurring excessive computational overhead. An adaptive fusion strategy further enhances flexibility, enabling the model to dynamically emphasize the most informative components of the input. In addition, we incorporate a more robust optimization objective that improves learning stability across varying sample difficulties and class distributions. Extensive evaluations on 30 UEA benchmark datasets and real-world vibration data demonstrate that FreSH consistently outperforms state-of-the-art methods in classification accuracy, while substantially reducing model size and efficiency.

cs.LG↗

arXiv2026

El Agente Gráfico: A Semantic Execution Runtime for Scientific Agents

Large language models (LLMs) can plan scientific workflows and generate code, but these capabilities do not specify how scientific state is validated, transferred and recorded across heterogeneous computational and experimental operations. Here we present El Agente Gráfico, a semantic execution runtime for scientific agents that uses typed execution graphs to enforce admissible scientific state transitions, record provenance and confine model judgement to explicit decision points. Using the same top-level LLM and task-specific rubrics on six university-level quantum chemistry exercises, El Agente Gráfico improved performance while reducing model cost by approximately 80% and wall-clock time by more than fourfold relative to our previous multi-agent architecture. Across ensemble spectroscopy and metal-organic framework (MOF) design, the runtime supported typed-state transfer, parallel execution and cross-session persistence. Open-ended studies of a disulfide redox mechanism and the vibrational hypothesis of olfaction required the agent to define the chemical scope and computational pathway of each investigation. A graph-construction skill distilled from these studies then enabled coding agents to author typed execution graphs for MOF literature mining. Together, these results show how typed execution graphs turn reusable procedural knowledge into scientific protocols that can be inspected, revised and transferred across computational and physical systems.

cs.AI↗

arXiv2026

Blockchain Empowered Trustworthy Agent Networks: Foundations, Taxonomy, and Future Directions

AI agents are evolving from isolated task executors into networked autonomous entities that can communicate, delegate tasks, invoke tools, access external knowledge, and participate in cross-platform service and economic workflows. This evolution gives rise to open agent networks, where heterogeneous agents owned by different stakeholders interact without naturally shared infrastructures for identity, authorization, auditability, reputation, or settlement. This survey and tutorial article reviews the literature over the period 1980--2026 on the evolution from classical multi-agent systems to open agent networks, with a particular focus on LLM-based autonomous agents, agent interoperability protocols, Internet-of-Agents infrastructures, and blockchain-enabled trust mechanisms. We first review this evolution and show how the trust boundary expands from individual execution to cross-agent, cross-platform, and cross-organizational interaction. We then identify a network-level trust crisis that cannot be fully addressed by single-agent safety mechanisms or closed multi-agent coordination techniques, and develop a five-dimensional taxonomy covering entity and capability trust, authorization and delegation trust, information and provenance trust, coordination and group-robustness trust, and accountability and settlement trust. Based on this taxonomy, we examine how blockchain can provide shared identity, verifiable authorization, tamper-evident provenance, auditable collaboration, incentive alignment, and value settlement for trustworthy agent networks. We further synthesize the mapping between agent-network risks, trust requirements, and blockchain-enabled mechanisms, and clarify the role of blockchain as a shared trust layer rather than a replacement for agent security, semantic verification, privacy protection, or robust reasoning.

cs.CR↗

arXiv2026

Combating Knowledge Corruption in Agent Systems: A Byzantine-Tolerant Secure Collaborative RAG Framework

While retrieval-augmented generation systems partially address the hallucination issues in large language models, it also introduces new vulnerabilities to knowledge corruption attacks. Adversaries exploit these vulnerabilities by poisoning documents provided by RAG system to manipulate LLM outputs. To counter this threat, we propose SecureCollaRAG, a Byzantine-tolerant collaborative RAG framework leveraging Multi-source Knowledge Validation Mechanism. Our approach enables agent system to securely verify document provenance through dynamic GNN-based credibility scoring, effectively preventing stealthy knowledge corruption attacks while preserving essential domain knowledge integrity. Through extensive evaluations and formal analysis, we demonstrate that SecureCollaRAG maintains robustness against attackers under non-IID data distributions.

cs.CR↗

arXiv2026

SubQuad: Near-Quadratic-Free Structure Inference with Distribution-Balanced Objectives in Adaptive Receptor framework

Comparative analysis of adaptive immune repertoires at population scale is hampered by two practical bottlenecks: the near-quadratic cost of pairwise affinity evaluations and dataset imbalances that obscure clinically important minority clonotypes. We introduce SubQuad, an end-to-end pipeline that addresses these challenges by combining antigen-aware, near-subquadratic retrieval with GPU-accelerated affinity kernels, learned multimodal fusion, and fairness-constrained clustering. The system employs compact MinHash prefiltering to sharply reduce candidate comparisons, a differentiable gating module that adaptively weights complementary alignment and embedding channels on a per-pair basis, and an automated calibration routine that enforces proportional representation of rare antigen-specific subgroups. On large viral and tumor repertoires SubQuad achieves measured gains in throughput and peak memory usage while preserving or improving recall@k, cluster purity, and subgroup equity. By co-designing indexing, similarity fusion, and equity-aware objectives, SubQuad offers a scalable, bias-aware platform for repertoire mining and downstream translational tasks such as vaccine target prioritization and biomarker discovery.

cs.LG↗