arXiv subjects

Lu Zhang

Publications and source records attributed to Lu Zhang.

At least 19 recordsLinked to original sources

arXiv2026

LIBERO-RECOVER: Beyond Task Success Towards Failure Recovery in Robotic Manipulation Models

Vision-Language-Action (VLA) or World Action (WAM) models have recently demonstrated remarkable performance in robotic manipulation. On LIBERO, SOTA method have achieved nearly 100\% success rates, seemingly suggesting that the models are ready for deployment in real world. However, near perfect performance on existing benchmarks can be misleading: success under ideal conditions does not imply real world robustness. Existing benchmarks primarily evaluate task completion from predefined initial states, while real world interactions inevitably involve failures such as failed grasps, collisions, and unintended object movements. A robot must therefore not only execute tasks successfully, but also recognize and recover from failures to continue the task. Yet this capability remains largely unmeasured, revealing a critical gap between benchmark performance and real world reliability. To address this gap, we introduce LIBERO-Recover Benchmark, a large scale benchmark for failure recovery in robotic manipulation. Built upon LIBERO, we collect real execution failures from SOTA embodied models and construct 1,000+ scenarios across four recovery levels: (1) Action Retry, (2) Action Adaptation, (3) Object State Recovery, and (4) Environmental Recovery. We evaluate four core capabilities: spatial understanding, object structure reasoning, interaction understanding, and topological reasoning. As the first large-scale benchmark for embodied failure recovery, LIBERO-Recover shifts evaluation from \emph{Can the robot succeed?''} to \emph{Can the robot recover after failure?''}, promoting robust and generalizable embodied agents. The project will be avaible in \textcolor{blue}{https://liulin815.github.io/LIBERO-Recovery/}.

cs.RO↗

arXiv2026

Planning-Coding Separation in LLM-Based Code Generation: An Empirical Study

Planning has been identified as an important factor for improved code generation with Large Language Models (LLMs). However, the interaction between planning capability and coding capability remains largely underexplored, especially when the two capabilities are from different LLMs. In this paper, we conduct the first comprehensive empirical study of planning-coding interaction in LLM-based code generation, where planning and coding are assigned to different LLMs (i.e., planning-coding separation). Specifically, we explore different planner-coder combinations to investigate the effectiveness for coders to utilize planning information from various planners and the impact of further training of coders on planner-coder combinations. Our results demonstrate that almost all planner-coder combinations are as competitive as using the same LLM for planning and coding, and the combination of a stronger planner and a stronger coder typically achieves improved effectiveness. Furthermore, using a large model as the planner paired with a much smaller model as the coder can sometimes be as competitive as using the large model as both the planner and the coder, and further training of the coder can typically boost various planner-coder combinations. These findings provide empirical evidence that planning and coding can be separated to a considerable extent in LLM-based code generation, which may offer a more flexible way to construct code-generation systems.

cs.AI↗

arXiv2026

Unlocking the Potential of Vision Foundation Models for Multi-Task Dense Prediction via Mixture of Task Adapters

Multi-task dense prediction seeks a single, coherent representation that supports simultaneously discriminative semantic recognition and regularized geometric/structural inference. For vision foundation models, pretrained semantic invariances conflict with the frequency and smoothness biases of dense objectives; joint fine-tuning and single-route adapters therefore compress competing gradients into one update channel, causing negative transfer and unstable cross-task agreement. We frame this as two-level adaptation: task-generic alignment that preserves priors, and task-aware specialization that resolves late-stage conflicts. Based on this principle, we propose D2MoE , a Mixture of Task Adapters that explicitly factorizes these roles within a frozen DINOv3 backbone. D2MoE introduces Task-Generic Low-Rank Adaptation (TG-LoRA) in attention projections throughout the network to provide a stable, shared adaptation channel, and a Task-Adaptive Mixture of Experts (TA-MoE) injected only at the final attention block of each stage to enact sparse, task-conditioned perturbations of token mixing in a single forward pass. A Semantic Prior Injector further couples multi-level task features with frozen foundation priors to reinforce semantic coherence while preserving boundary and structural fidelity. Experiments across diverse benchmarks substantiate that this structured factorization yields robust multi-task improvements with a small trainable footprint, establishing a principled and scalable route to prior-preserving, foundation-centric multi-task dense perception.

cs.CV↗

arXiv2026

SkillEvoLean: Mutation-enhanced skill evolution for Lean provers

Skill evolution offers a promising way to improve large language model agents without updating their parameters, but its use in formal theorem proving remains underexplored. Existing methods mainly target natural-language reasoning, improving skills by analyzing successful and failed trajectories and incrementally revising solving strategies. Although the Lean verifier provides reliable execution feedback, when all sampled trajectories fail, existing skill evolution methods lack successful trajectories from which to infer effective update directions. Furthermore, these methods also focus mainly on the root instruction file, thus underexploring the evolution of reference knowledge including mathematical concepts and proving techniques. To address these limitations, we propose a mutation-enhanced skill self-evolution framework for building skill-augmented Lean provers. The framework jointly evolves a high-level solving policy and its reference knowledge through progressive and mutation-based updates. Progressive evolution derives local improvements from successful and failed trajectories, while mutation is triggered when no complete proof can be generated, sampling mathematical concepts to produce and select new skill candidates under verifier feedback. We evaluate our method on MiniF2F, PutnamBench, the 2025 International Mathematical Olympiad (IMO 2025), and the 2026 USA Mathematical Olympiad (USAMO 2026). Under the same backbone model, trajectorysampling budget, and test-time compute, our method achieves proof success rates of 100.0%, 90.6%, 4/6, and 4/6, respectively, with GPT-5.5, outperforming the baseline methods. Further analysis shows that concept-guided mutation outperforms random-text-guided mutation by 6.9 and 8.2 percentage points on MiniF2F and PutnamBench, respectively, while solving one additional problem on both IMO 2025 and USAMO 2026.

cs.LG↗

arXiv2026

Spectral Invariant Fusion and Refinement for Domain Generalized Semantic Segmentation

Domain generalized semantic segmentation is challenging because it requires accurate dense prediction on unseen domains without any access to target-domain data during training. Existing methods improve transferability through data diversification, invariant learning, or direct use of foundation models, but they often struggle to simultaneously suppress domain-sensitive interference and preserve fine-grained spatial cues. In this work, we propose SIFR, a frequency-aware dual-stream framework that tackles domain shift at both the fusion and decoding stages. SIFR uses spectral modeling as a unified principle to regulate both heterogeneous prior interaction and dense decoder refinement. Instead of directly fusing frozen CLIP and DINOv2 features, SIFR performs spectral-guided invariant fusion to filter out domain-sensitive responses while preserving complementary semantic and structural information. In addition, SIFR incorporates a spectral decoder refinement module that stabilizes multi-scale decoder features through low-frequency statistics-based normalization and adaptive frequency selection, thereby reducing the propagation of residual domain bias in dense prediction. This design enables SIFR to jointly improve domain robustness and segmentation precision under severe distribution shifts. Extensive experiments on multiple DGSS benchmarks, including synthetic-to-real, real-to-real, and adverse-weather settings, demonstrate consistent improvements over previous state-of-the-art methods.

cs.CV↗

arXiv2026

FeatLens: Feature-Guided Dependency Retrieval for Repository-Level Code Generation

Retrieving task-relevant code dependencies is critical for repository-level code generation. Existing approaches obtain dependency context through similarity-based retrieval, repository graphs, or LLM-driven exploration. A central challenge is to identify the dispersed dependencies needed to implement a functional requirement, rather than merely retrieve broadly related code. This paper presents FeatLens, a feature-guided dependency retrieval approach for repository-level code generation. FeatLens first builds an offline index linking natural-language feature descriptions to function-level code entities. Given a task requirement and target function location, it selects relevant features and constructs dependency seeds. It then expands the seed graph and applies semantic and structural heuristics to select dependencies for inclusion in the generation context. Experiments on DevEval and EvoCodeBench show that FeatLens achieves the highest dependency recall among the evaluated baselines at each fixed retrieval budget, reaching DR@15 values of 0.527 and 0.529, respectively. On DevEval, the retrieved context supports the highest non-oracle Pass@1 with both DeepSeek-V3.2 and GPT-5-mini. These results are obtained without LLM calls in the online dependency retrieval stage.

cs.SE↗

arXiv2026

An Energy-based discontinuous Galerkin method for the Westervelt equation with absorbing boundary conditions

This paper develops and analyzes an energy-based discontinuous Galerkin method for the Westervelt equation in the acoustic-velocity-potential formulation with spatially varying material parameters and nonlinear absorbing boundary conditions. The semi-discrete scheme is constructed from the underlying energy balance and incorporates an additional consistent term to accommodate the strong damping operator within the discrete energy framework. Numerical fluxes are designed at both interelement and physical boundaries to provide a unified treatment of interior and boundary dissipation. For the nonlinear analysis, we study an auxiliary variable-coefficient linearized problem and establish stability and a priori error estimates, including the additional terms induced by the solution-dependent absorbing boundary flux. These estimates are combined with a fixed-point argument to prove existence, uniqueness, and convergence of the nonlinear semi-discrete approximation under suitable regularity, smallness, and nondegeneracy assumptions. Numerical experiments confirm the theoretical convergence results and illustrate the performance of the proposed method.

math.NA↗

arXiv2026

Magnetic Imaging with High Spatial Resolution Using a Single Niobium-Nitride-Based Three-Dimensional Nanobridge Junction

The nanobridge junctions (NBJs) allow the miniaturization of the Superconducting Quantum Interference Device (SQUID) and improve the spatial resolution of scanning SQUID microscopy. However, the SQUID loop diameter imposed a fundamental limit on the resolution. In this work, we proposed a single niobium-nitride-based three-dimensional nanobridge junction (NbN-3D-NBJ) alternative to the nanoSQUID for high-resolution magnetic imaging. The critical current of an NbN-3D-NBJ showed an nearly monotonic dependence on the external magnetic field, enabling nanoscale mapping of the magnetic field distribution by measuring the junction's critical current with an intrinsic noise level of 100 nT/Hz^1/2. Using a NbN-3D-NBJ probe, we successfully imaged an Abrikosov vortex in a niobium film with a minimal full width at half maxima (FWHM) of 530 nm, constrained by the distance from the junction to the probe tip edge. Furthermore, we resolved 80 nm spacings between the superconducting nanowires utilizing a second probe with an optimized junction-to-tip distance. Thus, the single 3D NbN NBJ probe has been demonstrated as an effective approach for magnetic imaging with ultra-high spatial resolution.

cond-mat.mes-hall↗

arXiv2026

3D-SceneQ: Empowering 3D LLM With Query-Guided Adaptive Pruning and Multi-Modal Feature Enhancement

Current 3D scene understanding pipelines typically concatenate vast numbers of 3D object tokens with text tokens and feed the resulting sequence to a Large Language Model (LLM). However, existing 3D-LLMs for scene understanding encounter two major limitations: (i) excessive inclusion of task-irrelevant object data, introducing noise that reduces reasoning accuracy and increases hallucinations; and (ii) reliance solely on point cloud data, which inherently lacks the rich semantic information available in complementary 2D modalities, such as color, material properties, texture, and high-level contextual relationships. To address these challenges, we introduce 3D-SceneQ, a 3D LLM that unifies adaptive token pruning with multimodal semantic enrichment, markedly advancing scene understanding, reasoning, and grounding. Specifically, we propose a Query-Guided Adaptive Pruning (QGAP) module that adaptively selects task-relevant objects based on the intended meaning of user instructions, leveraging a learnable latent query to retain critical information while effectively suppressing irrelevant noise. In addition, we introduce a Multi-modal Object-level Feature Enhancement (MOFE) module, which integrates 2D feature embeddings from pre-trained models into 3D representation to enrich object-level semantic information. Evaluations show that 3D-SceneQ outperforms current state-of-the-art methods across multiple benchmarks (e.g., +5.7 CIDEr on Scan2Cap and +17.0 EM@1 on SQA3D), demonstrating its remarkable capabilities in various 3D language reasoning tasks.

cs.CV↗

arXiv2026

Rivet: Utility-Preserving Context Control for Repository-Level Coding Agents

Repository-level code generation asks an agent to implement a task requirement inside an existing repository. To produce code that works in the repository, the agent needs repository evidence, such as target locations, helper definitions, caller usages, and configuration constraints, which determine how the code should be located, integrated, and executed. Because the context budget cannot hold the entire repository, existing retrieval, compression, and structure-aware methods select context by task similarity, context length, or structural proximity, but these criteria do not directly measure whether the selected context contains this evidence. As a result, the selected context can appear relevant to the task yet miss the evidence that the implementation depends on. We call this failure mode context utility dilution. In this paper, we propose Rivet, a context control framework that preserves repository evidence for coding agents under a fixed context budget. Rivet, first uses Signal Grounding to select task-related code anchors by fusing lexical, semantic, and structural signals. Utility Linking then expands these anchors through dependency, usage, and configuration relations to collect repository evidence. Finally, Budget Shaping packs the evidence by priority and, when the budget is tight, reduces snippets to signatures or summaries before dropping them. We evaluate Rivet, on two executable repository-level benchmarks, DevEval and RepoExec, with four backbone models, two context budgets (4K and 8K tokens), and eight baselines. Rivet, achieves the highest Pass@1 in all 16 settings and improves over the strongest baseline by 3.73--5.15 percentage points on average. Moreover, GPT-5-mini with Rivet, outperforms GPT-5 without Rivet,on both benchmarks while reducing inference cost by 73.3%--84.5%.

cs.SE↗

arXiv2026

Singularity Analysis of Laplace Boundary Integral Equations on Three-Dimensional Domains with Corners

Solutions to elliptic boundary value problems on three-dimensional nonsmooth domains exhibit singular behavior near edges and vertices. Using layer potential theory, we investigate boundary integral equations for the Laplace equation and characterize the leading singular exponents of the surface density on conical domains. For two canonical geometries, we derive explicit formulas for the two families of edge singular exponents on a dihedral wedge, and characterize the modewise vertex singular exponents on an axisymmetric circular cone through the zeros of associated Legendre functions and their derivatives. Building on these two prototypes and the dilation-invariance property, we formulate the associated operator pencil for general conical domains and develop a nonlinear eigenvalue solver. This solver employs the Nyström method with graded meshes, logarithmically corrected trapezoidal rules, and Brent's method to compute these singular exponents accurately and efficiently. Numerical examples for trihedral cones, regular m-hedral cones, and droplet-type cones verify the effectiveness of our approach and illustrate the interactions between edge and vertex singularities.

math.NA↗

arXiv2026

DisasterRoute: Evidence-Guided Dual-System Routing for Visual Disaster Question Answering

Visual question answering from low-altitude UAV imagery must deliver timely and accurate judgments to support disaster response under limited edge computing resources. Routine questions may be answered through direct decisions, whereas more challenging questions may benefit from step-by-step reasoning. However, existing research primarily focuses on improving models' visual understanding and generative reasoning capabilities, with limited exploration of efficient direct decision-making and its adaptive coordination with generative reasoning. We develop Disaster-S1 based on a vision-language model adapted to the disaster domain. Trained for option discrimination, it outputs a probability distribution over answer options in a single multimodal forward pass for fast decision-making. We further propose DisasterRoute, a dual-system routing framework that combines prediction confidence with disaster evidence to decide whether to accept the initial answer or invoke the disaster reasoning model DisasterVL for review. By allocating inference computation on demand, the framework retains the efficiency of direct decision-making and improves overall question-answering accuracy through selective review. Experiments on DisasterBench show that Disaster-S1 achieves superior accuracy with low latency, while DisasterRoute further improves accuracy through selective review at substantially lower latency than full reasoning.

cs.CV↗

arXiv2026

An exploration of plug-and-play flow matching for full-waveform inversion on breast imaging

Full-waveform inversion (FWI) reconstructs sound-speed fields from measured waveforms, by solving an optimization problem that minimizes the discrepancy between measured waveforms and those predicted by a numerical wave-propagation model. However, choosing suitable regularization for this optimization is challenging because it must stabilize the inversion while preserving relevant structures, prompting us to investigate whether priors learned from representative sound-speed images can improve reconstruction. In this work, we adapt plug-and-play flow matching (PnP-Flow) to FWI by alternating waveform-misfit updates with updates from a pretrained flow prior. We explore the effectiveness of PnP-Flow reconstruction by comparing it with total-variation (TV) regularization using noiseless transmission-only plane-wave data. Separate flow priors are trained on two circular-inclusion model families with constant and spatially varying backgrounds. We evaluate the method both on these in-distribution circle models and under substantial distribution shift by transferring the Gaussian-circle prior and reconstruction settings to breast phantom patches without retraining or retuning. Using the same number of waveform-misfit updates, the proposed PnP-Flow framework generally improves circle reconstruction accuracy and recovers breast lesion contrast more accurately than TV baselines.

eess.IV↗

arXiv2026

Reliability-Consistent Appearance Modeling for RGB-Infrared Multi-Object Tracking

Recent RGB-infrared multi-object tracking (MOT) methods have substantially improved cross-modal representation learning and multimodal information integration. However, most existing methods compress RGB and infrared evidence into a unified descriptor. When one modality is temporarily degraded, the tracker can no longer regulate its contribution separately; unreliable evidence may distort association and contaminate appearance memory. We therefore propose RCA-MOT, a reliability-consistent appearance modeling framework that preserves cross-modal complementarity while retaining post-representation evidence controllability. RGB-specific, infrared-specific, and fused identity representations are maintained as jointly accessible appearance states. For each detection, a simple training-free target-scene estimator produces branch-wise appearance reliability used across the two temporal roles of appearance: relative reliability regulates different identity branches, absolute reliability calibrates borderline candidates near the ReID acceptance boundary, and the same branch-wise reliability controls how strongly matched descriptors update historical appearance memory. The individual components are intentionally simple, focusing evaluation on reliability-controllable appearance organization rather than elaborate fusion or learned reliability-estimation modules. RCA-MOT sets new SOTA results on VT-MOT and UniRTL, improving the previous best HOTA by 5.402/6.075 points and IDF1 by 11.549/12.789 points, respectively, while post-hoc analyses further validate reliability as an informative control signal. Code will be released upon acceptance.

cs.CV↗

arXiv2026

State-Aware Generation of Deployable Websites from Visual Prototypes and Requirements

Generating deployable websites from visual prototypes and textual requirements is increasingly feasible with multimodal models, yet keeping generated projects consistent with their prototypes remains challenging. Prototypes specify layouts and visual elements, while requirements describe interaction workflows and deployment constraints. Existing UI-to-code methods and coding agents often treat layout reconstruction, interaction implementation, and project editing separately, allowing visual elements, files, workflows, dependencies, and render evidence to lose their correspondence across generation, verification, and update. We call this problem prototype-to-project state drift. We propose VisStateCoder, a state-aware framework that maintains these correspondences in a Project State Graph. The graph connects four types of state nodes through typed links and supports prototype grounding, graph-guided generation, and render-evidence-guided scoped updates that map failed checks to affected files while preserving files outside the update scope. We evaluate VisStateCoder on Vision2Web using five foundation models across Static Webpage, Interactive Frontend, and Full-Stack Website tasks. Averaged across models, VisStateCoder improves overall scores over the strongest baseline for each model by 26.1 and 22.6 points on Interactive Frontend and Full-Stack Website tasks, respectively. It also reduces verified-state drift, measured by checks that pass in one update round but fail later, compared with alternative state representations. An annotation study of 50 tasks finds that the constructed graphs achieve 95.6% edge-level F1.

cs.SE↗

arXiv2026

Do Modifications Made by Coding Agents Make Subsequent Software Maintenance Harder?

Coding agents increasingly maintain repositories over time, where each change becomes the starting point for subsequent tasks. Yet existing benchmarks either start tasks from developer-written code or cannot separate later-task difficulty from the effects of an agent's earlier changes. We ask whether agent-written code makes subsequent maintenance harder and whether the cost stems from inherited test failures or the code itself. We introduce SWE-Maintain, comprising 100 three-event maintenance episodes from 50 real-world Python repositories, with hidden tests certified across each sequence. Each later task is evaluated twice with the same agent and tests: from the developers' preceding patches (Developer-History) and from the agent's own changes (Agent-History). Across five coding systems, two backbone models, and 1,996 paired later events, Agent-History reduces later-task resolution by 5.6 percentage points on average. When success additionally requires all earlier tests to pass, the gap reaches 14.0 points. Two-thirds of Agent-History starting repositories already fail earlier tests, and only 1.9% of events starting from those states end with all accumulated tests passing. When the inherited code passes earlier tests and the next task is not already solved, the disadvantage disappears (+2.7 points; 95% CI [-0.5, +6.1]). These findings suggest that the main downstream cost of agent-written changes lies in inherited test failures rather than a measurable maintenance disadvantage in code that preserves earlier tested behavior. Evaluations should therefore include agent-produced history, and coding agents should preserve earlier behavior across changes.

cs.SE↗

arXiv2026

Boosting the photon bunching of classical light via a modified Fabry-Pérot cavity

We proposed to boost the photon bunching effect of classical light via a modified Fabry-Pérot cavity with one of the cavity mirrors being replaced by a spatial light modulator, through which multiple independent transverse modulations, each was to encode independently spatially correlated phase on the wave front of the light field, was cascaded in sequence in the longitudinal propagation direction in the cavity. The two-photon bunching peak and the three-photon bunching peak of the generated superbunching light field were measured to be $36.9\pm2.0$ and $\rm (2.46\pm0.12)\times10^{3}$, respectively, which, as compared to those of thermal light, were improved by one-order and three-order of magnitude, respectively. Correspondingly, the ghost image visibility reconstructed with the superbunching light was demonstrated experimentally to be improved by $\sim 40$ times as compared to that with the pseudo-thermal light. Such classical superbunching light field is of essential importance not only for fundamental physics but also for potential applications such as correlated imaging, ghost interference and nonlinear optics.

physics.optics↗

arXiv2026

Route What Remains: A Meta-Modal Agent for Missing-Modality Candidate Reranking in Recommender Systems

Missing-modality recommenders usually reconstruct absent representations, although the observed evidence may not determine the missing content. We formulate candidate reranking as budgeted sequential evidence acquisition. A policy queries text, image, and interaction-graph tools, incorporates \texttt{Null} returns into its observation history, and sparsely rescores a retrieved candidate pool. Our \textbf{Meta-Modal Agent} (MMA) uses PPO to optimize terminal NDCG and tool cost without explicit access to the route-availability mask or target identity. When only one evidence route is available, MMA-Auto improves NDCG@10 by $10.0$\% over the strongest completion baseline and by $9.5$\% over a fixed router with the same Llama scorer. It obtains the highest result in all nine reported combinations of dataset and available route against these comparators. MMA-Auto also reduces failed calls by 17.8 percentage points and uses 1.1 fewer turns than the fixed router. On the fixed candidate pools produced by full-catalog retrieval, MMA-Auto improves NDCG@10 by $19.7$\%. These results associate adaptive evidence routing with improved reranking under severe, constructed missingness. The code is available at: https://anonymous.4open.science/r/WSDM2027-MMA-C381.

cs.IR↗