arXiv subjects

Li Zeng

Publications and source records attributed to Li Zeng.

At least 37 records · Page 2Linked to original sources

arXiv2026

Beyond Bellman: High-Order Generator Regression for Continuous-Time Policy Evaluation

We study finite-horizon continuous-time policy evaluation from discrete closed-loop trajectories under time-inhomogeneous dynamics. The target value surface solves a backward parabolic equation, but the Bellman baseline obtained from one-step recursion is only first-order in the grid width. We estimate the time-dependent generator from multi-step transitions using moment-matching coefficients that cancel lower-order truncation terms, and combine the resulting surrogate with backward regression. The main theory gives an end-to-end decomposition into generator misspecification, projection error, pooling bias, finite-sample error, and start-up error, together with a decision-frequency regime map explaining when higher-order gains should be visible. Across calibration studies, four-scale benchmarks, feature and start-up ablations, and gain-mismatch stress tests, the second-order estimator consistently improves on the Bellman baseline and remains stable in the regime where the theory predicts visible gains. These results position high-order generator regression as an interpretable continuous-time policy-evaluation method with a clear operating region.

stat.ML↗

arXiv2026

Fisher Decorator: Refining Flow Policy via a Local Transport Map

Recent advances in flow-based offline reinforcement learning (RL) have achieved strong performance by parameterizing policies via flow matching. However, they still face critical trade-offs among expressiveness, optimality, and efficiency. In particular, existing flow policies interpret the $L_2$ regularization as an upper bound of the 2-Wasserstein distance ($W_2$), which can be problematic in offline settings. This issue stems from a fundamental geometric mismatch: the behavioral policy manifold is inherently anisotropic, whereas the $L_2$ (or upper bound of $W_2$) regularization is isotropic and density-insensitive, leading to systematically misaligned optimization directions. To address this, we revisit offline RL from a geometric perspective and show that policy refinement can be formulated as a local transport map: an initial flow policy augmented by a residual displacement. By analyzing the induced density transformation, we derive a local quadratic approximation of the KL-constrained objective governed by the Fisher information matrix, enabling a tractable anisotropic optimization formulation. By leveraging the score function embedded in the flow velocity, we obtain a corresponding quadratic constraint for efficient optimization. Our results reveal that the optimality gap in prior methods arises from their isotropic approximation. In contrast, our framework achieves a controllable approximation error within a provable neighborhood of the optimal solution. Extensive experiments demonstrate state-of-the-art performance across diverse offline RL benchmarks. See project page: https://github.com/ARC0127/Fisher-Decorator.

cs.LG↗

arXiv2026

Exoplanet Orbital Distribution around FGK Sun-like Host Stars II: a valley in the orbital semi-major axis distribution of sub-Neptunes

More than one hundred years ago, physics has been revolutionized when people realized that electronic orbitals, or electromagnetic interactions in general, are quantized. Now, in this study, we are presenting evidence of quantization of planet orbits around stars. Confining a wave in spatial dimensions "quantizes" its wave number. Therefore, this study points to the evidence of the existence of long-range standing waves in the proto-planetary disks. Such waves, although being on a much larger scale of few tens of AU, have already been found by ALMA observation-so called ring-like structure. Now we see that it may exist within 1 AU, and may exert its effect on the existence and distribution of planets within this distance range to the host star. Careful analysis has been carried out to compare the results of different surveys.

astro-ph.EP↗

arXiv2026

Service Function Chain Routing in LEO Networks Using Shortest-Path Delay Statistical Stability

Low Earth orbit (LEO) satellite constellations have become a critical enabler for global coverage, utilizing numerous satellites orbiting Earth at high speeds. By decomposing complex network services into lightweight service functions, network function virtualization (NFV) transforms global network services into diverse service function chains (SFCs), coordinated by resource-constrained LEOs. However, the dynamic topology of satellite networks, marked by highly variable inter-satellite link delays, poses significant challenges for designing efficient routing strategies that ensure reliable and low-latency communication. Many existing routing methods suffer from poor scalability and degraded performance, limiting their practical implementation. To address these challenges, this paper proposes a novel SFC routing approach that leverages the statistical properties of network link states to mitigate instability caused by instantaneous modeling in dynamic satellite networks. Through comprehensive simulations on end-to-end shortest-path propagation delays in LEO networks, we identify and validate the statistical stability of multi-hop routes. Building on this insight, we introduce the Stability-Aware Multi-Stage Graph Routing (SA-MSGR) algorithm, which incorporates pre-computed average delays into a multi-stage graph optimization framework. Extensive simulations demonstrate the superior performance of SA-MSGR, achieving significantly lower and more predictable end-to-end SFC delays compared to representative baseline strategies.

cs.NI↗

arXiv2026

Unseen Cost of Space Computing: Quantifying LEO Battery Aging via Physics-Driven Modeling

Low Earth Orbit (LEO) satellite constellations in the 6G era are evolving into intelligent in-orbit computational platforms, forming Space Computing Power Networks (SCPNs) to deliver global-scale computing services. However, the intensive computation within SCPN incurs a significant ``unseen cost'': the frequent charge-discharge cycles accelerate the physical degradation of satellites' life-limiting and high-cost batteries, thereby threatening the long-term operational viability of such a system. Existing approaches, often relying on indirect metrics like Depth of Discharge (DoD) and neglecting the complex, nonlinear degradation process of battery aging, fail to accurately quantify this cost. To address this, we introduce a high-fidelity, physics-driven model that quantitatively links computational workload parameters to the nonlinear battery degradation. Building on this model, we formulate a degradation-aware scheduling problem and analyze heuristic policies across different energy regimes. Simulations reveal that the optimal strategy should be adaptive: in solar-rich conditions, a myopic policy maximizing instantaneous solar utilization is superior, whereas under energy scarcity, a reactive policy leveraging real-time battery state significantly extends lifetime.

eess.SP↗

arXiv2026

Optimizing p-spin models through hypergraph neural networks and deep reinforcement learning

p-spin glasses, characterized by frustrated many-body interactions beyond the conventional pairwise case (p>2), are prototypical disordered systems whose ground-state search is NP-hard and computationally prohibitive for large instances. Solving this problem is not only fundamental for understanding high-order disorder, structural glasses, and topological phases, but also central to a wide spectrum of hard combinatorial optimization tasks. Despite decades of progress, there still lacks an efficient and scalable solver for generic large-scale p-spin models. Here we introduce PLANCK, a physics-inspired deep reinforcement learning framework built on hypergraph neural networks. PLANCK directly optimizes arbitrary high-order interactions, and systematically exploits gauge symmetry throughout both training and inference. Trained exclusively on small synthetic instances, PLANCK exhibits strong zero-shot generalization to systems orders of magnitude larger, and consistently outperforms state-of-the-art thermal annealing methods across all tested structural topologies and coupling distributions. Moreover, without any modification, PLANCK achieves near-optimal solutions for a broad class of NP-hard combinatorial problems, including random k-XORSAT, hypergraph max-cut, and conventional max-cut. The presented framework provides a physics-inspired algorithmic paradigm that bridges statistical mechanics and reinforcement learning. The symmetry-aware design not only advances the tractable frontiers of high-order disordered systems, but also opens a promising avenue for machine-learning-based solvers to tackle previously intractable combinatorial optimization challenges.

cond-mat.dis-nn↗

arXiv2026

Beyond Static Question Banks: Dynamic Knowledge Expansion via LLM-Automated Graph Construction and Adaptive Generation

Personalized education systems increasingly rely on structured knowledge representations to support adaptive learning and question generation. However, existing approaches face two fundamental limitations. First, constructing and maintaining knowledge graphs for educational content largely depends on manual curation, resulting in high cost and poor scalability. Second, most personalized education systems lack effective support for state-aware and systematic reasoning over learners' knowledge, and therefore rely on static question banks with limited adaptability. To address these challenges, this paper proposes a Generative GraphRAG framework for automated knowledge modeling and personalized exercise generation. It consists of two core modules. The first module, Automated Hierarchical Knowledge Graph Constructor (Auto-HKG), leverages LLMs to automatically construct hierarchical knowledge graphs that capture structured concepts and their semantic relations from educational resources. The second module, Cognitive GraphRAG (CG-RAG), performs graph-based reasoning over a learner mastery graph and combines it with retrieval-augmented generation to produce personalized exercises that adapt to individual learning states. The proposed framework has been deployed in real-world educational scenarios, where it receives favorable user feedback, suggesting its potential to support practical personalized education systems.

cs.CY↗

arXiv2025

CrowdLLM: Building LLM-Based Digital Populations Augmented with Generative Models

The emergence of large language models (LLMs) has sparked much interest in creating LLM-based digital populations that can be applied to many applications such as social simulation, crowdsourcing, marketing, and recommendation systems. A digital population can reduce the cost of recruiting human participants and alleviate many concerns related to human subject study. However, research has found that most of the existing works rely solely on LLMs and could not sufficiently capture the accuracy and diversity of a real human population. To address this limitation, we propose CrowdLLM that integrates pretrained LLMs and generative models to enhance the diversity and fidelity of the digital population. We conduct theoretical analysis of CrowdLLM regarding its great potential in creating cost-effective, sufficiently representative, scalable digital populations that can match the quality of a real crowd. Comprehensive experiments are also conducted across multiple domains (e.g., crowdsourcing, voting, user rating) and simulation studies which demonstrate that CrowdLLM achieves promising performance in both accuracy and distributional fidelity to human data.

cs.MA↗

arXiv2025

RoME: Domain-Robust Mixture-of-Experts for MILP Solution Prediction across Domains

Mixed-Integer Linear Programming (MILP) is a fundamental and powerful framework for modeling complex optimization problems across diverse domains. Recently, learning-based methods have shown great promise in accelerating MILP solvers by predicting high-quality solutions. However, most existing approaches are developed and evaluated in single-domain settings, limiting their ability to generalize to unseen problem distributions. This limitation poses a major obstacle to building scalable and general-purpose learning-based solvers. To address this challenge, we introduce RoME, a domain-Robust Mixture-of-Experts framework for predicting MILP solutions across domains. RoME dynamically routes problem instances to specialized experts based on learned task embeddings. The model is trained using a two-level distributionally robust optimization strategy: inter-domain to mitigate global shifts across domains, and intra-domain to enhance local robustness by introducing perturbations on task embeddings. We reveal that cross-domain training not only enhances the model's generalization capability to unseen domains but also improves performance within each individual domain by encouraging the model to capture more general intrinsic combinatorial patterns. Specifically, a single RoME model trained on three domains achieves an average improvement of 67.7% then evaluated on five diverse domains. We further test the pretrained model on MIPLIB in a zero-shot setting, demonstrating its ability to deliver measurable performance gains on challenging real-world instances where existing learning-based approaches often struggle to generalize.

cs.LG↗

arXiv2025

EDBench: Large-Scale Electron Density Data for Molecular Modeling

Existing molecular machine learning force fields (MLFFs) generally focus on the learning of atoms, molecules, and simple quantum chemical properties (such as energy and force), but ignore the importance of electron density (ED) $ρ(r)$ in accurately understanding molecular force fields (MFFs). ED describes the probability of finding electrons at specific locations around atoms or molecules, which uniquely determines all ground state properties (such as energy, molecular structure, etc.) of interactive multi-particle systems according to the Hohenberg-Kohn theorem. However, the calculation of ED relies on the time-consuming first-principles density functional theory (DFT) which leads to the lack of large-scale ED data and limits its application in MLFFs. In this paper, we introduce EDBench, a large-scale, high-quality dataset of ED designed to advance learning-based research at the electronic scale. Built upon the PCQM4Mv2, EDBench provides accurate ED data, covering 3.3 million molecules. To comprehensively evaluate the ability of models to understand and utilize electronic information, we design a suite of ED-centric benchmark tasks spanning prediction, retrieval, and generation. Our evaluation on several state-of-the-art methods demonstrates that learning from EDBench is not only feasible but also achieves high accuracy. Moreover, we show that learning-based method can efficiently calculate ED with comparable precision while significantly reducing the computational cost relative to traditional DFT calculations. All data and benchmarks from EDBench will be freely available, laying a robust foundation for ED-driven drug discovery and materials science.

physics.chem-ph↗

arXiv2025

Bounded KRnet and its applications to density estimation and approximation

In this paper, we develop an invertible mapping, called B-KRnet, on a bounded domain and apply it to density estimation/approximation for data or the solutions of PDEs such as the Fokker-Planck equation and the Keller-Segel equation. Similar to KRnet, B-KRnet consists of a series of coupling layers with progressively fewer active transformation dimensions, inspired by the triangular structure of the Knothe-Rosenblatt (KR) rearrangement. The main difference between B-KRnet and KRnet is that B-KRnet is defined on a hypercube while KRnet is defined on the whole space, in other words, a new mechanism is introduced in B-KRnet to maintain the exact invertibility. Using B-KRnet as a transport map, we obtain an explicit probability density function (PDF) model that corresponds to the pushforward of a base (uniform) distribution on the hypercube. It can be directly applied to density estimation when only data are available. By coupling KRnet and B-KRnet, we define a deep generative model on a high-dimensional domain where some dimensions are bounded and other dimensions are unbounded. A typical case is the solution of the stationary kinetic Fokker-Planck equation, which is a PDF of position and momentum. Based on B-KRnet, we develop an adaptive learning approach to approximate partial differential equations whose solutions are PDFs or can be treated as PDFs. A variety of numerical experiments is presented to demonstrate the effectiveness of B-KRnet.

cs.LG↗

arXiv2025

Taylor$\unicode{x2013}$Aris dispersion of active particles in oscillatory channel flow

Mass dispersion in oscillatory flows is intimately linked to various environmental and biological processes, offering a distinct contrast to dispersion in steady flows due to the periodic expansion and contraction of particle patches. In this study, we investigate the Taylor$\unicode{x2013}$Aris dispersion of active particles in laminar oscillatory flows between parallel plates. Two complementary approaches are employed: a two-time-variable expansion of the Smoluchowski equation is used to facilitate Aris' method of moments for the preasymptotic dispersion, while the generalised Taylor dispersion theory is extended to capture phase-dependent periodic drift and dispersivity in the long-time asymptotic limit. Applying both frameworks, we find that spherical non-gyrotactic swimmers can exhibit greater or lesser diffusivity than passive solutes in purely oscillatory flows, depending on the oscillation frequency. This behaviour arise primarily from the disruption of cross-streamline migration governed by Jeffery orbits. When a steady component is superimposed, oscillation induces a non-monotonic dual effect on diffusivity. We further examine two well-studied shear-related accumulation mechanisms, arising from gyrotaxis and elongation. Although these accumulation effects are less pronounced than in steady flows due to flow unsteadiness, gyrotactic swimmers respond more effectively to the unsteady shear profile, significantly altering their drift and dispersivity. This work offers new insights into the dispersion of active particles in oscillatory flows and also provides a foundation for studying periodic active dispersion beyond the oscillatory flow, such as periodic variations in shape and swimming speed.

physics.flu-dyn↗

arXiv2025

Missing data in non-stationary multivariate time series from digital studies in Psychiatry

Mobile technology (e.g., mobile phones and wearable devices) provides scalable methods for collecting physiological and behavioral biomarkers in patients' naturalistic settings, as well as opportunities for therapeutic advancements and scientific discoveries regarding the etiology of psychiatric illness. Continuous data collection through mobile devices generates highly complex data: entangled multivariate time series of outcomes, exposures, and covariates. Missing data is a pervasive problem in biomedical and social science research, and Ecological Momentary Assessment (EMA) data in psychiatric research is no exception. However, the complex data structure of multivariate time series and their non-stationary nature make missing data a major challenge for proper inference. Additional historical information included in time series analyses exacerbates the issue of missing data and also introduces problems for confounding adjustment. The majority of existing imputation methods are either designed for stationary time series or for longitudinal data with limited follow-up periods. The limited work on non-stationary time series either focuses on missing exogenous information or ignores the complex temporal dependence among outcomes, exposures, and covariates. We propose a Monte Carlo Expectation Maximization algorithm for the state space model (MCEM-SSM) to effectively handle missing data in non-stationary entangled multivariate time series. We demonstrate the method's advantages over other widely used missing data imputation strategies through simulations of both stationary and non-stationary time series, subject to various missing mechanisms. Finally, we apply the MCEM-SSM to a multi-year smartphone observational study of bipolar and schizophrenia patients to investigate the association between digital social connectivity and negative mood.

stat.ME↗

arXiv2025

WindVE: Collaborative CPU-NPU Vector Embedding

Retrieval-Augmented Generation is a technology that enhances large language models by integrating information retrieval. In the industry, inference services based on LLMs are highly sensitive to cost-performance ratio, prompting the need for improving hardware resource utilization in the inference service. Specifically, vector embedding and retrieval processes take up to 20% of the total latency. Therefore, optimizing the utilization of computational resources in vector embeddings is crucial for enhancing the cost-performance ratio of inference processes, which in turn boosts their product competitiveness.In this paper, we analyze the deployment costs of vector embedding technology in inference services, propose a theoretical formula, and determine through the mathematical expression that increasing the capacity to process concurrent queries is the key to reducing the deployment costs of vector embeddings. Therefore, in this paper, we focus on improving the product's capability to process concurrent queries. To optimize concurrency without sacrificing performance, we have designed a queue manager that adeptly offloads CPU peak queries. This manager utilizes a linear regression model to ascertain the optimal queue depths, a critical parameter that significantly influences the efficacy of the system. We further develop a system named WindVE that uses a CPU-NPU heterogeneous architecture to offload peak concurrent queries, which leverages the performance differences between the two processors to effectively manage traffic surges. Through experiments, we compare WindVE to the state-of-the-art vector embedding framework FlagEmbedding, and achieve a concurrency level up to 22.3% higher than the scheme without offloading.

cs.DC↗

arXiv2025

SLO-Aware Scheduling for Large Language Model Inferences

Large language models (LLMs) have revolutionized applications such as code completion, chatbots, and online classification. To elevate user experiences, service level objectives (SLOs) serve as crucial benchmarks for assessing inference services capabilities. In practice, an inference service processes multiple types of tasks, each with its own distinct SLO. To ensure satisfactory user experiences, each request's distinct SLOs should be considered in scheduling. However, existing designs lack this consideration, leading to insufficient hardware utility and suboptimal performance. This paper analyzes scenarios to process tasks with varying SLOs, and introduces a simulated annealing-based scheduler to decide request priority sequence based on a request's SLO, input lengths, and possible output lengths. As the first specialized scheduler for multi-SLO scenarios, this work improves SLO attainment by up to 5x and reduces average latency by 31.6% on Python-Code-23k-ShareGPT and ShareGPT_Vicuna_unfiltered datasets, compared to current state-of-the-art framework vLLM and a new framework LMDeploy.

cs.DC↗

arXiv2025

High-Efficiency Plasma-Based Compressor for Ultrafast Soft X-ray Free-Electron Lasers

The generation of intense, femtosecond-scale X-ray pulses is crucial for probing matter under extreme temporal and field conditions. Current chirped-pulse amplification (CPA) techniques in free-electron lasers (FELs), however, face efficiency limitations in the soft X-ray regime due to the inherent constraints of conventional optical compressors. To address this challenge, we propose a high-efficiency plasma-based compressor utilizing highly ionized noble gas plasma. Exploiting strong refractive index dispersion near ionic resonances, this scheme achieves over 70% transmission efficiency around 5.2 nm, and is extendable to other highly charged ions for operation across the soft X-ray to vacuum ultraviolet range. Simulations demonstrate that a 25 fs FEL pulse can be compressed to 1.4 fs with peak power boosted to over 100 GW, while maintaining high energy throughput. This approach overcomes the long-standing efficiency bottleneck of soft X-ray CPA and opens a scalable path toward compact, high-brightness attosecond FEL sources.

physics.acc-ph↗

arXiv2025

DocMEdit: Towards Document-Level Model Editing

Model editing aims to correct errors and outdated knowledge in the Large language models (LLMs) with minimal cost. Prior research has proposed a variety of datasets to assess the effectiveness of these model editing methods. However, most existing datasets only require models to output short phrases or sentences, overlooks the widespread existence of document-level tasks in the real world, raising doubts about their practical usability. Aimed at addressing this limitation and promoting the application of model editing in real-world scenarios, we propose the task of document-level model editing. To tackle such challenges and enhance model capabilities in practical settings, we introduce \benchmarkname, a dataset focused on document-level model editing, characterized by document-level inputs and outputs, extrapolative, and multiple facts within a single edit. We propose a series of evaluation metrics and experiments. The results show that the difficulties in document-level model editing pose challenges for existing model editing methods.

cs.CL↗

arXiv2025

Demonstration of Direct-amplification Enabled Harmonic Generation in an Ultraviolet Free-Electron Laser

We report the experimental demonstration of direct-amplification enabled harmonic generation in an ultraviolet free-electron laser (FEL) driven by a low-intensity seed laser. By employing a versatile undulator configuration that enables seed amplification and harmonic generation within a unified setup, we achieved over 100-fold energy gain of the seed and observed exponential growth at the second harmonic. The results demonstrate that a sufficiently long modulator can not only amplify a weak seed but also induce strong energy modulation of the electron beam, enabling efficient harmonic bunching. This method markedly relaxes the power requirements on external seed lasers and presents a viable route toward high-repetition-rate, fully coherent FELs

physics.acc-ph↗