Searcharxiv⌕ Search

arXiv subjects

Xiaoyang Wang

Publications and source records attributed to Xiaoyang Wang.

At least 19 recordsLinked to original sources
arXiv2026

An Empirical Study of Harness Design for Coding Agents

Coding harnesses shape how autonomous coding agents translate model capabilities into long-horizon software-engineering performance, yet existing work typically evaluates harnesses as monolithic systems, leaving the effectiveness of individual components unclear. To enable component-level comparisons, we study this question with a lightweight coding harness whose execution loop is fixed while three components are varied: planning, action space, and context management. Across four models evaluated on SWE-Bench Verified and Terminal-Bench 2.1, we evaluate 176 matched settings spanning five context-management strategies, four context-window budgets, and targeted ablations of planning and action space. We find that: (1) Context management becomes increasingly valuable as the context-window budget tightens, with most of its benefit coming from preventing context-overflow failures. (2) Staging rule-based elision before LLM-based summarization provides the strongest overall efficiency among the context-management strategies, whereas making elided content recoverable adds machinery that models rarely use and yields no accuracy gain. (3) Planning shifts from an accuracy scaffold for weaker models to a cost saver for stronger models, with little change in accuracy. (4) Predefined tools improve performance for models with weaker bash proficiency, whereas bash-capable models can operate effectively with a bash-only interface and achieve substantially lower cost, especially on command-line-centric tasks. Trajectory-level analysis explains these effects: context management extends execution trajectories without substantially altering agent behavior, planning changes where trajectories stop, and the action space changes the granularity at which code is written. These findings inform model- and budget-aware harness design and provide a modular framework for evaluating future harness components.

cs.AI↗
arXiv2026

Recursive Self-Improvement LLM Agents for Inverter Dynamic Model Identification

This is a position paper. We demonstrate that recursive self-improvement (RSI) large language model (LLM) agents are a natural search engine for dynamic model identification of inverter-based resources (IBRs) whose internal controls are often proprietary and hidden from grid operators. White-box models provide physical transparency but require vendor disclosure; black-box models avoid this requirement but sacrifice interpretability; and existing grey-box approaches, including sparse and symbolic regression, are poorly suited to discovering feedback control architectures or incorporating control-engineering priors. Our position is that this gap can be alleviated by (1) restricting the search space to a typed vocabulary of standard control modules, including PI controllers, phase-locked loops (PLLs), low-pass filters, etc., composed under block-diagram grammar rules, and (2) using an RSI LLM agent to perform program search over candidate block-diagram models, guided by measured frequency-domain admittance data at the point of common coupling (PCC), while fitting the free parameters of each candidate by nonlinear least squares. We instantiate this position by adapting ThetaEvolve, an open-source program-evolution framework supporting in-context evolution and test-time learning, to inverter model discovery. In a proof-of-concept study on a grid-following (GFL) inverter benchmark, the RSI loop reduces the normalized root mean square error (NRMSE) of a naive open-loop model from 0.470 to 0.0435 and identifies a 15-module closed-loop structure that closely resembles the hidden ground-truth GFL controller.

eess.SY↗
arXiv2026

Joint Mitigation of Algorithmic and Physical Errors in Noisy Hamiltonian Simulation

Product-formula Hamiltonian simulation is naturally suited to near-term quantum processors, but its accuracy is set by two competing errors: finite-step Trotter bias and physical hardware noise. We introduce a joint extrapolation strategy that ties the tunable per-layer noise strength to the Trotter step size, $λ(s)=c(sT)^{p+1}$ for a $p$-th order product formula. Along this one-dimensional path, the leading physical-noise and Trotter corrections over the full evolution both enter at order $s^p$ and can be canceled by a single Richardson extrapolation. Building on a previously established finite-order Baker--Campbell--Hausdorff truncation bound, we derive a provably commutator-scaling resource guarantee for the joint extrapolation. For local Hamiltonians with local Lindbladian noise, the protocol mitigates physical noise together with Trotter error with only a constant asymptotic overhead relative to noiseless Trotter extrapolation, provided the required noise strengths lie above the intrinsic device-noise floor. We experimentally demonstrate the protocol for Ising dynamics on a superconducting quantum computer using learned noise amplification. Complementary 100-qubit Sparse Pauli Dynamics (SPD) simulations achieve comparable accuracy to two-dimensional Richardson extrapolation with fewer circuit settings.

quant-ph↗
arXiv2026

LightNav-0: Eliciting VLM Spatial Intelligence for Generalist Embodied Navigation

Embodied navigation requires agents to translate heterogeneous goals and visual observations into actions across tasks, environments, and robot embodiments. Modern vision-language models (VLMs) already encode spatial priors for visual grounding, spatial reasoning, and pointing, but these capabilities are rarely elicited directly for robot control. Existing navigation systems instead rely on task- or embodiment-specific components, fragmenting perception, reasoning, and action while offering limited generalization. Here we present LightNav-0, a compact generalist embodied navigation model that elicits the spatial intelligence of a pretrained VLM and aligns it with navigation, without task-specific prediction heads. LightNav-0 represents diverse navigation tasks through a unified token interface: dual-channel pointing expresses task-, scene-, and embodiment-agnostic spatial intent, while a residual vector-quantized action tokenizer maps this intent to precise, embodiment-specific trajectories. Together with temporally aware visual history compression, ER mid-training, supervised fine-tuning, and reinforcement learning, this formulation supports instruction following, open-vocabulary object navigation, and visual tracking within a single model. The navigation training corpus spans 2K+ scenes and 4K+ hours of embodied navigation data. LightNav-ER, the embodied-reasoning checkpoint used to initialize LightNav-0, attains the highest complete-set average across 8 embodied-reasoning benchmarks, while LightNav-0 achieves state-of-the-art monocular success rates across all 10 public navigation simulation settings. Real-world evaluations further demonstrate zero-shot generalization across robot embodiments, diverse scenes, and static and dynamic targets. These results establish compact VLMs as a unified and transferable backbone for generalist embodied navigation.

cs.RO↗
arXiv2026

SPARK: Skeleton-Guided Reasoning Synthesis from Large-Scale Scientific Literature

Scientific reasoning remains challenging for open-source models, largely due to the lack of high-quality scientific reasoning data. Existing datasets are often dominated by factual recall or formulaic problem solving, with limited emphasis on mechanism understanding, evidence-grounded reasoning, and hypothesis evaluation. To address this, we introduce SPARK (Scientific Paper Abstracted Reasoning sKeleton), a paper-oriented synthesis framework built on Sci-Base, a large-scale corpus of research papers spanning 10 scientific disciplines. Instead of directly converting papers into question-answer pairs, SPARK treats the claim-evidence-derivation structure of a paper as the fundamental unit of reasoning synthesis. Specifically, SPARK (1) distills each paper into a compact reasoning skeleton capturing its central claims and supporting evidence, enabling self-contained question generation, and (2) synthesizes reasoning tasks from four scientific perspectives: mechanistic reasoning, hypothesis falsification, quantitative derivation, and boundary calibration. A final consistency verification stage further removes unsupported or contradictory outputs. Using this framework, we construct Spark-234K, a scientific reasoning dataset with substantially higher difficulty and diversity than existing resources. Experiments show that Spark-234K consistently outperforms existing scientific reasoning datasets while achieving stronger performance with significantly fewer training samples.

cs.AI↗
arXiv2026

Performance guarantees of light-cone variational quantum algorithms for the maximum cut problem

Variational quantum algorithms (VQAs) are promising to demonstrate the advantage of near-term quantum computing over classical computing in practical applications, such as the maximum cut (MaxCut) problem. However, current VQAs such as the quantum approximate optimization algorithm (QAOA) have lower performance guarantees compared to the best-known classical algorithm, and suffer from hard optimization processes due to the barren plateau problem. We propose a light-cone VQA by choosing an optimal gate sequence of the standard VQAs, which enables a significant improvement in solution accuracy while avoiding the barren plateau problem. Specifically, we prove that the light-cone VQA with one round achieves an approximation ratio of 0.7926 for the MaxCut problem in the worst case of $3$-regular graphs, which is higher than that of the 3-round QAOA, and can be further improved to 0.8333 by an angle-relaxation procedure. Finally, our numerical results indicate an exponential speed-up in finding the exact solution using the light-cone VQA compared with the classical algorithm. Using IBM's quantum devices, we demonstrate that the single-round light-cone VQA exceeds the known classical hardness threshold in both 72- and 148-qubit demonstrations, whereas $p$-round $\text{QAOA}$ with $p=1,2,3$ does not in the latter one. Our work highlights a promising route towards solving classically hard problems on practical quantum devices.

quant-ph↗
arXiv2026

Scientific Graphics Program Synthesis via Dual Self-Consistency Reinforcement Learning

Graphics Program Synthesis is pivotal for interpreting and editing visual data, effectively facilitating the reverse-engineering of static visuals into editable TikZ code. While TikZ is the de facto standard for scientific schematics due to its programmatic flexibility, its requirement for rigorous spatial precision presents a significant challenge for Multimodal Large Language Models. Progress is currently stifled by two primary gaps: (1) Data Quality Gap: existing image-TikZ corpora often lack strict executability and reliable visual alignment; (2) Evaluation Gap: a lack of benchmarks for both structural and visual fidelity. To address these, we present a closed-loop framework featuring: SciTikZ-230K, a large-scale, high-quality dataset from our Execution-Centric Data Engine covering 11 diverse scientific disciplines; SciTikZ-Bench, a multifaceted benchmark spanning from basic geometric constructs to intricate hierarchical schematics to evaluate both visual fidelity and structural logic. To further broaden the scope of visual-code optimization methodology, we introduce a novel Dual Self-Consistency Reinforcement Learning optimization paradigm, which utilizes Round-Trip Verification to penalize degenerate code and boost overall self-consistency. Empowered by these, our trained model SciTikZer-8B achieves state-of-the-art performance, consistently outperforming proprietary giants like Gemini-2.5-Pro and massive models like Qwen3-VL-235B-A22B-Instruct.

cs.CV↗
arXiv2026

Efficient Quantum Simulations of Yang-Mills theory with Maximal-tree Gauge

We develop a quantum algorithmic framework for the efficient simulation of Yang--Mills theories, including the $\mathrm{SU}(3)$ gauge theory in Quantum Chromodynamics (QCD). The framework uses maximal-tree gauge in terms of gauge field variables that removes all local gauge redundancies. In the resulting gauge-fixed formulation and digitization in the field-amplitude basis, we show that Hamiltonian time evolution admits an efficient implementation based on quantum singular value transformation (QSVT). We derive upper bounds on the total number of qubits and gate complexity, finding polynomial scaling with the inverse simulation precision $1/\varepsilon_s$, lattice volume $\mathcal{V}$, gauge coupling $g$, and target energy scale $E$. Our results provide a rigorous complexity-theoretic demonstration that non-Abelian Yang--Mills theories can be simulated efficiently on quantum computers, paving the way toward first-principles quantum simulations of non-perturbative QCD dynamics.

hep-lat↗
arXiv2026

Lindbladian Simulation with Commutator Bounds

Trotter decomposition provides a simple approach to simulating open quantum systems by decomposing the Lindbladian into a sum of individual terms. While it is established that Trotter errors in Hamiltonian simulation depend on nested commutators of the summands, such a relationship remains poorly understood for Lindbladian dynamics. In this Letter, we derive commutator-based Trotter error bounds for Lindbladian simulation, yielding an $O(\sqrt{N})$ scaling in the number of Trotter steps for locally interacting systems on $N$ sites. When estimating observable averages, we apply Richardson extrapolation to achieve polylogarithmic precision while maintaining the commutator scaling. To bound the extrapolation remainder, we develop a general truncation bound for the Baker-Campbell-Hausdorff expansion that bypasses common convergence issues in physically relevant systems. For local Lindbladians, our results demonstrate that the Trotter-based methods outperform prior simulation techniques in system-size scaling while requiring only $O(1)$ ancillas. Numerical simulations further validate the predicted system-size and precision scaling.

quant-ph↗
arXiv2026

Trace Only What You Need: Structure-Aware On-Demand Hypergraph Memory for Long-Document Question Answering

Long-document question answering (QA) requires large language models (LLMs) to reason over evidence scattered across lengthy documents, where answers often depend on event order, section-level context, and cross-part evidence connections. Although retrieval-augmented generation (RAG) reduces the input context by retrieving relevant evidence, existing structured RAG methods still face three limitations: costly query-agnostic knowledge organization, insufficient use of original document structure, and no reuse of historical reasoning experience. To address these limitations, we propose DocTrace, a multi-agent RAG framework for long-document QA that supports query-triggered knowledge organization, document-structure-aware and experience-guided reasoning. DocTrace preserves document hierarchy with a lightweight document structural tree index, constructs agent-shared hypergraph-structured working memory on demand during reasoning, and stores successful reasoning plans in graph-structured experience memory for future reuse, enabling adaptive exploration across related long-document questions. Experiments on four long-document QA datasets show that DocTrace outperforms the strongest baseline, ComoRAG, with average relative gains of 16.91% in F1 and 15.50% in EM on open-form QA benchmarks, and by up to 20.67% and 23.78%, respectively, on NarrativeQA.

cs.CL↗
arXiv2026

SkillZip: Contract-Preserving Graph Compression for Scalable Agent Skill Libraries

Large Language Models (LLMs) increasingly act as agents whose procedural knowledge is stored in reusable skill packages and loaded at inference time. As skill libraries grow, a central challenge is to expose the smallest sufficient executable context under a limited context budget. Existing systems struggle to reuse routines below the whole-skill level, preserve procedural contracts during compression, keep compressed routines executable and expandable, and update the compressed library as skills evolve. These challenges reveal a unit mismatch: skills are retrieved as packages, compressed as text, and converted into execution graphs only after retrieval, whereas reliable reuse requires a contract-bearing procedural unit. We propose SkillZip, an execution-aware procedural abstraction framework that performs contract-preserving compression over section-level graphs. SkillZip rewrites recurring contract-valid motifs into reversible ported macros while preserving boundary signatures, dependency closure, verifier reachability, and source-level expansion. At inference time, it hydrates a compact, dependency-closed context and expands macros only when required. ReZip further integrates new skills and revises risky macros using execution evidence. Comprehensive experiments1 on technical and embodied agent benchmarks show SkillZip consistently outperforms the strongest baseline by up to 12.2 points, while achieving a 3.46x compression ratio with 99.2% dependency preservation and 98.7% verifier reachability. Scaling analyses further confirm robust retrieval across skill libraries ranging from 200 to 100K skills.

cs.CL↗
arXiv2026

Electricity Demand and Grid Impacts of AI Data Centers: Challenges and Prospects

The rapid growth of artificial intelligence (AI) is driving an unprecedented increase in the electricity demand of AI data centers, raising emerging challenges for electric power grids. Understanding the characteristics of AI data center loads and their interactions with the grid is therefore critical for ensuring both reliable power system operation and sustainable AI development. This paper provides a comprehensive review and vision of this evolving landscape. Specifically, this paper (i) presents an overview of AI data center infrastructure and its key components, (ii) examines the key characteristics and patterns of electricity demand across the stages of model preparation, training, fine-tuning, and inference, (iii) analyzes the critical challenges that AI data center loads pose to power systems across three interrelated timescales, including long-term planning and interconnection, short-term operation and electricity markets, and real-time dynamics and stability, and (iv) discusses potential solutions from the perspectives of the grid, AI data centers, and AI end-users to address these challenges. By synthesizing current knowledge and outlining future directions, this review aims to guide research and development in support of the joint advancement of AI data centers and power systems toward reliable, efficient, and sustainable operation.

eess.SY↗
arXiv2026

HyperAgent: Planning and Acting over Tool-Schema Hypergraphs for Tool-Use LLM Agents

Large language model (LLM) agents increasingly rely on external tools to complete complex real-world tasks. However, reliable tool-use planning remains challenging due to the limitations of implicit reasoning and the evolving nature of real-world execution environments. Existing tool-use agents typically rely on LLMs to infer tool compositions from textual descriptions, which can lead to inefficient exploration and unreliable execution in complex tasks. To address these challenges, we model tool relations at the schema level and construct a directed Tool--Schema Hypergraph, in which tools are represented as hyperedges from their required input-schema nodes to their output-schema nodes. Furthermore, we propose HyperAgent, a Tool--Schema Hypergraph-guided framework for dynamic planning and execution. Given a task, HyperAgent first extracts a task-relevant tool context graph and uses it to guide the construction of a schema-aware Task DAG. During execution, HyperAgent dynamically realizes each subtask by constructing a state-conditioned tool support graph through deficit-oriented expansion, which identifies unresolved requirements and retrieves supporting producer tools according to the current agent state. Experiments on AppWorld demonstrate that HyperAgent improves task completion performance while reducing redundant API calls, LLM interactions, and token consumption compared with existing agent baselines.

cs.AI↗
arXiv2026

A Unified Framework for Hybrid Grid-Forming and Grid-Following Inverter Control

This paper proposes a novel unified control framework for achieving hybrid grid-forming (GFM) and grid-following (GFL) inverter operation by integrating dispatchable virtual oscillator control with reference-following synchronization. The proposed inverter control method supports multiple operating modes within a unified structure, including voltage- and frequency-following (PQ mode), voltage-forming and frequency-following (PV mode), voltage-following and frequency-forming (Qf mode), voltage- and frequency-forming (Vf mode), and a hybrid mode with mixed GFM and GFL behaviors. In particular, the proposed method achieves smooth pre-synchronization and enables seamless transitions across a spectrum of inverter operating modes by tuning a small set of continuous control parameters, rather than relying on discrete controller switching. This framework provides a flexible and physically interpretable approach for adapting inverter dynamics to varying grid conditions and operational requirements. The small-signal stability and input-output frequency-domain characteristics are further analyzed under different control parameter settings. The effectiveness and robustness of the proposed unified control method are demonstrated through extensive electromagnetic transient (EMT) simulations and hardware-in-the-loop (HIL) experiments.

eess.SY↗
arXiv2026

Simultaneous Estimation of Partial-Transpose Moments with Active Memory Independent of the Moment Order

We study the simultaneous estimation of partial-transpose moments $p_j(ρ_{AB})=\mathrm{Tr}[(ρ_{AB}^{T_B})^j]$, $j=2,\ldots,K$, of an unknown bipartite $n$-qubit state from independent copies under an explicit active-memory constraint. We give a sequential qubit-reuse realization of the partial-transpose permutation that uses at most $2n+1$ active qubits, independent of $K$, and estimates all moments $p_2,\ldots,p_K$ to uniform additive error $ε$ with total copy complexity $O(K\log K/ε^2)$. We also prove two converse bounds. First, any uniformly accurate simultaneous estimator requires $Ω(K/ε^2)$ copies in the worst case. Second, the same scaling holds on an explicit isospectral two-qubit negative-partial-transpose (NPT) family whose ordinary moments are constant while the partial-transpose moments vary. These results characterize the copy complexity of the partial-transpose moment hierarchy up to a logarithmic factor and extend simultaneous nonlinear-functional estimation from ordinary state powers to partial-transpose spectral data under active quantum memory independent of the target moment order.

quant-ph↗
arXiv2026

Weaving Multi-Source Evidence for Biomedical Reasoning: The BioMedHop Benchmark and BioWeave Framework

Biomedical question answering (QA) increasingly requires reasoning over interacting entities, where supporting evidence is scattered across biomedical knowledge graphs, literature documents, and web-accessible resources. However, existing biomedical QA benchmarks mainly focus on exam-style knowledge, literature comprehension, or short-range multi-hop inference, leaving source-conditioned graph reasoning and evidence topology construction underexplored. To fill this gap, we introduce BioMedHop, a multi-source graph-grounded benchmark for evaluating biomedical reasoning over structured evidence topologies. BioMedHop contains 10,045 instances across KG, document, web, and hybrid evidence settings, covering shared-neighbor matching, intersection reasoning, path-based reasoning, and counting, with option-based, open-ended, and numeric count renderings. To support this benchmark, we further propose BioWeave, a source-aware reasoning framework that retrieves biomedical KG paths, gathers supporting clues from documents and web sources, assembles them into a unified evidence graph, and verifies answers through entity-level evidence support. Comprehensive experiments show that BioWeave achieves the best overall performance among compared methods on BioMedHop, outperforming the strong hybrid baseline ToG-2 by 10.5% in the overall average. Moreover, BioWeave consistently improves different LLM backbones and enables smaller models, such as Qwen3-4B, to achieve reasoning performance comparable to GPT-4-Turbo.

cs.CL↗
arXiv2026

Selection Integrity for LLM Graph Memory: An Accumulability Criterion for Information-Flow-Blind Retrieval

Agent memory is moving to graphs, and the provenance defenses now being built for it all check one thing: the provenance of the records an agent retrieves. We show that this entire class of defense is blind by construction. A long-term graph memory runs a global selection step over writable graph structure, so structure that an untrusted principal writes changes \emph{which} authenticated facts are selected while the cited evidence stays fully authenticated; faithful information-flow control (IFC), checking the provenance of what the reader uses (all of it authenticated), makes the byte-identical decision to no defense at all, across document-QA substrates and real multi-session agent memory. In the most consequential instance, a no-source structural write silently misdirects $28$ irreversible ledger transfers over $499$ live actions: faithful IFC permits every one, and \authselect\ prevents every one. We then characterize exactly which memories are exposed: a selector admits the channel when its structural term can reallocate an $Ω(1)$ share of top-$k$ membership past a selected fact's margin. Personalized PageRank can, since a sourceless write reroutes conserved random-walk mass; a content-fixed reranker cannot, and Graphiti's node-distance, which leans on structure \emph{more} than PageRank does, stays immune. Reallocatability, not reliance, is the predictor. We prove the immune case in general and the open case under a chokepoint condition we verify. Closing the channel forces any provenance defense to recompute selection on the authenticated subgraph, which is what \authselect\ does, at zero over-block and $2$--$3\%$ latency.

cs.CR↗
arXiv2026

The Consistency Illusion: How Multi-Agent Debate Hides Reasoning Misalignment

Multi-agent LLM systems for medical question answering often treat consensus as a reliability signal: if multiple agents agree on an answer, it is presumed trustworthy. However, answer-level consensus does not entail reasoning-level alignment. We introduce CARA (Cross-Agent Reasoning Alignment), a family of automated metrics that measure whether agents who agree on an answer also agree on the reasoning. Applying CARA to a standard debate system on two medical QA benchmarks, MedQA-USMLE and MedThink-Bench, we identify the consistency illusion: a failure mode where debate reduces detectable contradictions between agents while simultaneously decreasing the semantic similarity of their reasoning chains; agents appear to agree more but reason less consistently. To improve this misalignment, we propose the Grounded Debate Protocol (GDP), a prompt-level intervention that requires agents to commit to named medical facts and take explicit stances on other agents' claims. GDP produces large, consistent alignment improvements, with Cohen's d ranging from +1.43 to +1.99, across two datasets and two backbone models, without adding LLM calls or modifying system architecture. Our results motivate cross-agent reasoning alignment as a quantity to audit alongside accuracy in safety-critical domains.

cs.MA↗