Search arXiv⌕ Search

arXiv subjects

Lei Zhu

Publications and source records attributed to Lei Zhu.

At least 73 records · Page 4Linked to original sources
arXiv2026

ActDistill: General Action-Guided Self-Derived Distillation for Efficient Vision-Language-Action Models

Recent Vision-Language-Action (VLA) models have shown impressive flexibility and generalization, yet their deployment in robotic manipulation remains limited by heavy computational overhead and inference latency. In this work, we present ActDistill, a general action-guided self-derived distillation framework that transfers the action prediction capability of any existing VLA model to a lightweight counterpart. Unlike previous efficiency strategies that primarily emphasize vision-language correlations, ActDistill leverages action priors to guide knowledge transfer and model compression, achieving action-oriented efficiency for VLA models. Specifically, we employ a well-trained VLA model as the teacher and introduce a graph-structured encapsulation strategy to explicitly model the hierarchical evolution of action prediction. The student model, derived from the graph-encapsulated teacher, is further equipped with a dynamic router that adaptively selects computation paths based on action prediction demands, guided by hierarchical graph-informed supervision to ensure smooth and efficient evolution. During inference, graph-related auxiliary components are removed, allowing the student to execute only dynamically routed layers and predict high-precision actions with minimal computation and latency. Experiments on embodied benchmarks demonstrate that ActDistill achieves comparable or superior performance to full-scale VLA models while reducing computation by over 50% with up to 1.67 times speedup, thereby establishing a general paradigm toward efficient embodied intelligence.

cs.CV↗
arXiv2026

RADA: Region-Aware Dual-encoder Auxiliary learning for Barely-supervised Medical Image Segmentation

Deep learning has greatly advanced medical image segmentation, but its success relies heavily on fully supervised learning, which requires dense annotations that are costly and time-consuming for 3D volumetric scans. Barely-supervised learning reduces annotation burden by using only a few labeled slices per volume. Existing methods typically propagate sparse annotations to unlabeled slices through geometric continuity to generate pseudo-labels, but this strategy lacks semantic understanding, often resulting in low-quality pseudo-labels. Furthermore, medical image segmentation is inherently a pixel-level visual understanding task, where accuracy fundamentally depends on the quality of local, fine-grained visual features. Inspired by this, we propose RADA, a novel Region-Aware Dual-encoder Auxiliary learning pipeline which introduces a dual-encoder framework pre-trained on Alpha-CLIP to extract fine-grained, region-specific visual features from the original images and limited annotations. The framework combines image-level fine-grained visual features with text-level semantic guidance, providing region-aware semantic supervision that bridges image-level semantics and pixel-level segmentation. Integrated into a triple-view training framework, RADA achieves SOTA performance under extremely sparse annotation settings on LA2018, KiTS19 and LiTS, demonstrating robust generalization across diverse datasets.

cs.CV↗
arXiv2026

DA-PTQ: Drift-Aware Post-Training Quantization for Efficient Vision-Language-Action Models

Vision-Language-Action models (VLAs) have demonstrated strong potential for embodied AI, yet their deployment on resource-limited robots remains challenging due to high memory and computational demands. While Post-Training Quantization (PTQ) provides an efficient solution, directly applying PTQ to VLAs often results in severe performance degradation during sequential control. We identify temporal error accumulation as a key factor, where quantization perturbations at the vision-language-to-action interface are progressively amplified, leading to kinematic drift in executed trajectories. To address this issue, we propose Drift-Aware Post-Training Quantization (DA-PTQ), which formulates quantization as a drift-aware optimization problem over sequential decision processes. DA-PTQ consists of two components: (1) Cross-Space Representation Compensation, which mitigates structured distortions between multimodal representations and action space to improve action consistency, and (2) Motion-Driven Mixed-Precision Allocation, which assigns bit-widths by minimizing trajectory-level motion errors. Extensive experiments show that DA-PTQ significantly reduces kinematic drift and achieves comparable performance to full-precision models under low-bit settings, enabling practical deployment of VLAs on resource-limited robotic platforms.

cs.RO↗
arXiv2026

High-resolution scanning fluorescence imaging through scattering via speckle replica alignment and variance computation

Fluorescence imaging is an essential diagnostic tool in many fields, but diffraction-limited optical imaging at depth is limited by scattering. Here, we present a method based on multiple random illuminations, combined with a computational framework that retrieves high-resolution images by aligning local speckle replicas and computing their pixel-wise variance. We demonstrate its versatility in two regimes: linear wide-field one-photon (1P) fluorescence imaging and nonlinear two-photon (2P) fluorescence imaging where the object is excited by a scanned speckle field and detected with a single-pixel detector. This approach outperforms standard autocorrelation techniques in terms of resolution and convergence.

physics.optics↗
arXiv2026

Diffusion-based Probabilistic Air Quality Forecasting with Mechanistic Insight

Current operational air quality forecasts are computationally expensive, sensitive to errors in physics and emissions, and often neglect weather-related uncertainty. To address these limitations, we present AirFusion, a hybrid, diffusion-based framework that synergistically integrates knowledge from chemical transport models with real-world observational constraints to enable accurate and efficient probabilistic regional air quality prediction. We apply AirFusion to generate operational 6-day, 30-member ensemble forecasts of surface ozone across China, initialized with observations and driven by ensemble weather forecasts. AirFusion outperforms existing operational benchmarks, achieving substantially lower forecast errors against surface measurements, while also providing ensemble-based diagnostics that explicitly quantify the impacts of weather uncertainty on air quality predictability. Moreover, AirFusion can rapidly adapt to evolving emissions through fine-tuning with only one month of recent observations. These attributes establish AirFusion as a powerful and extensible framework for next-generation probabilistic air quality forecasting, with clear potential for application to other pollutants and regions.

physics.ao-ph↗
arXiv2026

Monolithic integration of diverse crystalline thin films on diamond for near-junction thermal management

The pursuit of extreme miniaturization and high power in 6G RF front-ends has cast thermal dissipation as the central challenge. Here, we have demonstrated the monolithic integration of functionally distinct single-crystal thin films, including \b{eta}-Ga2O3, Si, GaN, and LiTaO3, onto a single diamond substrate using a multi-step transfer printing technique. Focusing on the critical \b{eta}-Ga2O3/diamond interface, we achieve an exceptional interfacial thermal conductance (ITC) of 149 MW m-2 K-1 through ultra-high vacuum (UHV) annealing, creating an atomically sharp interface featuring covalent bonding. Vibrational electron energy-loss spectroscopy (EELS) analysis combining with molecular dynamics (MD) simulations reveal that distinctive interfacial phonon modes at the \b{eta}-Ga2O3/diamond heterointerface dominate ultrahigh ITC. We experimentally demonstrate that by improving the ITC, the thermal resistance (Rth) of a diamond-based \b{eta}-Ga2O3 MOSFET is driven to a record-low value of 1.58 K mm W-1, underscoring the critical role of interface engineering in near-junction thermal management for diamond-integrated devices. This work demonstrates a scalable, diamond-based monolithic integration platform designed to solve the near-junction thermal challenges in high-power RF front-ends.

cond-mat.mtrl-sci↗
arXiv2026

How Long Can Unified Multimodal Models Generate Images Reliably? Taming Long-Horizon Interleaved Image Generation via Context Curation

Unified multimodal models hold the promise of generating extensive, interleaved narratives, weaving text and imagery into coherent long-form stories. However, current systems suffer from a critical reliability gap: as sequences grow, generation quality rapidly collapses. In this work, we investigate the mechanism behind this failure and argue that it is distinct from standard long-context challenges. We reveal that in generation, accumulated visual history acts as a source of active pollution, a decay governed specifically by the number of image events rather than raw token count. We identify a structural vulnerability where dense visual tokens overwhelm the attention mechanism, creating noise that distorts future synthesis. Guided by these mechanistic insights, we propose UniLongGen, a training-free inference strategy that prioritizes safe conditioning over total recall. Instead of retaining all history, UniLongGen dynamically curates the model's memory, identifying and discarding interfering visual signals based on the model's own internal relevance rankings. Extensive experiments demonstrate that this active forgetting approach is essential for stability: UniLongGen significantly outperforms baselines in long-horizon fidelity and consistency, while simultaneously reducing memory footprint and inference time.

cs.CV↗
arXiv2026

Reason Like a Radiologist: Chain-of-Thought and Reinforcement Learning for Verifiable Report Generation

Radiology report generation is critical for efficiency but current models lack the structured reasoning of experts, hindering clinical trust and explainability by failing to link visual findings to precise anatomical locations. This paper introduces BoxMed-RL, a groundbreaking unified training framework for generating spatially verifiable and explainable radiology reports. Built on a large vision-language model, BoxMed-RL revolutionizes report generation through two integrated phases: (1) In the Pretraining Phase, we refine the model via medical concept learning, using Chain-of-Thought supervision to internalize the radiologist-like workflow, followed by spatially verifiable reinforcement, which applies reinforcement learning to align medical findings with bounding boxes. (2) In the Downstream Adapter Phase, we freeze the pretrained weights and train a downstream adapter to ensure fluent and clinically credible reports. This framework precisely mimics radiologists' workflow, compelling the model to connect high-level medical concepts with definitive anatomical evidence. Extensive experiments on public datasets demonstrate that BoxMed-RL achieves an average 7% improvement in both METEOR and ROUGE-L metrics compared to state-of-the-art methods. An average 5% improvement in large language model-based metrics further underscores BoxMed-RL's robustness in generating high-quality radiology reports.

cs.AI↗
arXiv2026

Non-Markovian Long-Horizon Robot Manipulation via Keyframe Chaining

Existing Vision-Language-Action (VLA) models often struggle to generalize to long-horizon tasks due to their heavy reliance on immediate observations. While recent studies incorporate retrieval mechanisms or extend context windows to handle procedural tasks, they often struggle to capture Non-Markovian dependencies, where optimal actions rely solely on specific past states rather than the current observation. To address this, we introduce Keyframe-Chaining VLA, a framework that extracts and links key historical frames to model long-horizon dependencies. Specifically, we propose an automatic keyframe selector that learns a discriminative embedding space, effectively identifying distinct state transitions. To capture task-critical information, we design a progress-aware query mechanism that dynamically retrieves historical frames based on their temporal relevance to the current execution phase. These selected keyframes are integrated into the VLA as interleaved visual tokens, explicitly grounding the policy in the long-horizon temporal context. Finally, we introduce a suite of four Non-Markovian manipulation tasks built upon the ManiSkill simulator to measure task success rates. Experimental results demonstrate that our method achieves superior performance, effectively tackling robot manipulation tasks characterized by long-horizon temporal dependencies. Code is available at https://github.com/cytoplastm/KC-VLA.

cs.RO↗
arXiv2026

Self-Correcting VLA: Online Action Refinement via Sparse World Imagination

Standard vision-language-action (VLA) models rely on fitting statistical data priors, limiting their robust understanding of underlying physical dynamics. Reinforcement learning enhances physical grounding through exploration yet typically relies on external reward signals that remain isolated from the agent's internal states. World action models have emerged as a promising paradigm that integrates imagination and control to enable predictive planning. However, they rely on implicit context modeling, lacking explicit mechanisms for self-improvement. To solve these problems, we propose Self-Correcting VLA (SC-VLA), which achieve self-improvement by intrinsically guiding action refinement through sparse imagination. We first design sparse world imagination by integrating auxiliary predictive heads to forecast current task progress and future trajectory trends, thereby constraining the policy to encode short-term physical evolution. Then we introduce the online action refinement module to reshape progress-dependent dense rewards, adjusting trajectory orientation based on the predicted sparse future states. Evaluations on challenging robot manipulation tasks from simulation benchmarks and real-world settings demonstrate that SC-VLA achieve state-of-the-art performance, yielding the highest task throughput with 16% fewer steps and a 9% higher success rate than the best-performing baselines, alongside a 14% gain in real-world experiments. Code is available at https://github.com/Kisaragi0/SC-VLA.

cs.RO↗
arXiv2026

PosterReward: Unlocking Accurate Evaluation for High-Quality Graphic Design Generation

Recent advancements in the text-rendering capabilities of image generation models have made the end-to-end creation of graphic design content, such as posters, increasingly feasible. However, existing reward models fall short of accurately assessing design quality, as they primarily focus on global image aesthetics while overlooking the critical dimensions of typography and layout. Furthermore, the scarcity of domain-specific preference data remains a significant bottleneck, which limits the further development of graphic design evaluation and generation. To bridge this gap, we introduce an automated pipeline to construct a high-quality dataset of 70k poster preferences by leveraging the consensus of multiple Multi-modal Large Language Models (MLLMs) to simulate human-like judgment. Utilizing this dataset, we develop PosterReward, a reward model specifically designed for high-precision poster assessment through a cascaded, multi-stage training strategy. We also provide multiple variants of the model to cater to different application scenarios. Finally, we introduce PosterRewardBench and PosterBench to evaluate the performance of existing reward models in poster assessment and the generation capabilities of current text-to-image models in poster creation, respectively.

cs.GR↗
arXiv2026

Linear filament and nested cluster evolution tomography (LANCET) I. Capture the evolution of dense gas in 14-parsec filament G316.8

A dynamic view of mass assembly is essential for understanding the formation of massive stars and clusters. Interpreting evolutionary diagnostics from Galactic-wide surveys, however, requires careful control of distance and environmental variations. The G316.8 filament provides an ideal laboratory: a 14-pc nearly linear structure composed of three contiguous subregions with comparable molecular gas reservoirs (~10,000 $M_\odot$ each) but spanning a clear evolutionary sequence from an infrared dark cloud (young) through a massive young stellar object (intermediate) to an HII region (evolved). As part of the Linear filament and nested cluster evolution tomography (LANCET) project, we mapped the full filament with the Atacama Compact Array at 1.3 mm, achieving 0.08 pc resolution over 17.1 pc$^2$. Combined with Herschel and APEX/ArTéMiS data, we derived high-resolution temperature and column-density maps. We quantify structural evolution using dense-fragment statistics, column-density PDFs, and $Δ$-variance analysis. From young to evolved regions, the maximum fragment mass increases from 8 to 490 $M_\odot$, while the dense-gas mass fraction ($>0.5$ g cm$^{-2}$) rises from 0.4% to 9.6%. The N-PDF develops a secondary power-law tail and the $Δ$-variance slope becomes progressively shallower, indicating ongoing assembly of dense sub-parsec structures. Our further ALMA 12m continuum and spectral line data will extend this dynamic scenarios down to 800 AU scale.

astro-ph.GA↗
arXiv2026

MoESD: Unveil Speculative Decoding's Potential for Accelerating Sparse MoE

Large Language Models (LLMs) have achieved remarkable success across many applications, with Mixture of Experts (MoE) models demonstrating great potential. Compared to traditional dense models, MoEs achieve better performance with less computation. Speculative decoding (SD) is a widely used technique to accelerate LLM inference without accuracy loss, but it has been considered efficient only for dense models. In this work, we first demonstrate that, under medium batch sizes, MoE surprisingly benefits more from SD than dense models. Furthermore, as MoE becomes sparser -- the prevailing trend in MoE designs -- the batch size range where SD acceleration is expected to be effective becomes broader. To quantitatively understand tradeoffs involved in SD, we develop a reliable modeling based on theoretical analyses. While current SD research primarily focuses on improving acceptance rates of algorithms, changes in workload and model architecture can still lead to degraded SD acceleration even with high acceptance rates. To address this limitation, we introduce a new metric 'target efficiency' that characterizes these effects, thus helping researchers identify system bottlenecks and understand SD acceleration more comprehensively. For scenarios like private serving, this work unveils a new perspective to speed up MoE inference, where existing solutions struggle. Experiments on different GPUs show up to 2.29x speedup for Qwen2-57B-A14B at medium batch sizes and validate our theoretical predictions.

cs.LG↗
arXiv2026

MOTIF: Learning Action Motifs for Few-shot Cross-Embodiment Transfer

While vision-language-action (VLA) models have advanced generalist robotic learning, cross-embodiment transfer remains challenging due to kinematic heterogeneity and the high cost of collecting sufficient real-world demonstrations to support fine-tuning. Existing cross-embodiment policies typically rely on shared-private architectures, which suffer from limited capacity of private parameters and lack explicit adaptation mechanisms. To address these limitations, we introduce MOTIF for efficient few-shot cross-embodiment transfer that decouples embodiment-agnostic spatiotemporal patterns, termed action motifs, from heterogeneous action data. Specifically, MOTIF first learns unified motifs via vector quantization with progress-aware alignment and embodiment adversarial constraints to ensure temporal and cross-embodiment consistency. We then design a lightweight predictor that predicts these motifs from real-time inputs to guide a flow-matching policy, fusing them with robot-specific states to enable action generation on new embodiments. Evaluations across both simulation and real-world environments validate the superiority of MOTIF, which significantly outperforms strong baselines in few-shot transfer scenarios by 6.5% in simulation and 43.7% in real-world settings. Code is available at https://github.com/buduz/MOTIF.

cs.RO↗
arXiv2026

PosterOmni: Generalized Artistic Poster Creation via Task Distillation and Unified Reward Feedback

Image-to-poster generation is a high-demand task requiring not only local adjustments but also high-level design understanding. Models must generate text, layout, style, and visual elements while preserving semantic fidelity and aesthetic coherence. The process spans two regimes: local editing, where ID-driven generation, rescaling, filling, and extending must preserve concrete visual entities; and global creation, where layout- and style-driven tasks rely on understanding abstract design concepts. These intertwined demands make image-to-poster a multi-dimensional process coupling entity-preserving editing with concept-driven creation under image-prompt control. To address these challenges, we propose PosterOmni, a generalized artistic poster creation framework that unlocks the potential of a base edit model for multi-task image-to-poster generation. PosterOmni integrates the two regimes, namely local editing and global creation, within a single system through an efficient data-distillation-reward pipeline: (i) constructing multi-scenario image-to-poster datasets covering six task types across entity-based and concept-based creation; (ii) distilling knowledge between local and global experts for supervised fine-tuning; and (iii) applying unified PosterOmni Reward Feedback to jointly align visual entity-preserving and aesthetic preference across all tasks. Additionally, we establish PosterOmni-Bench, a unified benchmark for evaluating both local editing and global creation. Extensive experiments show that PosterOmni significantly enhances reference adherence, global composition quality, and aesthetic harmony, outperforming all open-source baselines and even surpassing several proprietary systems.

cs.CV↗
arXiv2026

Occlusion-Aware Consistent Model Predictive Control for Robot Navigation in Occluded Obstacle-Dense Environments

Ensuring safety and motion consistency for robot navigation in occluded, obstacle-dense environments is a critical challenge. In this context, this study presents an occlusion-aware Consistent Model Predictive Control (CMPC) strategy. To account for the occluded obstacles, it incorporates adjustable risk regions that represent their potential future locations. Subsequently, dynamic risk boundary constraints are developed online to enhance safety. Based on these constraints, the CMPC constructs multiple locally optimal trajectory branches (each tailored to different risk regions) to strike a balance between safety and performance. A shared consensus segment is generated to ensure smooth transitions between branches without significant velocity fluctuations, preserving motion consistency. To facilitate high computational efficiency and ensure coordination across local trajectories, we use the alternating direction method of multipliers (ADMM) to decompose the CMPC into manageable sub-problems for parallel solving. The proposed strategy is validated through simulations and real-world experiments on an Ackermann-steering robot platform. The results demonstrate the effectiveness of the proposed CMPC strategy through comparisons with baseline approaches in occluded, obstacle-dense environments.

cs.RO↗
arXiv2026

PEAR: Pixel-aligned Expressive humAn mesh Recovery

Reconstructing detailed 3D human meshes from a single in-the-wild image remains a fundamental challenge in computer vision. Existing SMPLX-based methods often suffer from slow inference, produce only coarse body poses, and exhibit misalignments or unnatural artifacts in fine-grained regions such as the face and hands. These issues make current approaches difficult to apply to downstream tasks. To address these challenges, we propose PEAR-a fast and robust framework for pixel-aligned expressive human mesh recovery. PEAR explicitly tackles three major limitations of existing methods: slow inference, inaccurate localization of fine-grained human pose details, and insufficient facial expression capture. Specifically, to enable real-time SMPLX parameter inference, we depart from prior designs that rely on high resolution inputs or multi-branch architectures. Instead, we adopt a clean and unified ViT-based model capable of recovering coarse 3D human geometry. To compensate for the loss of fine-grained details caused by this simplified architecture, we introduce pixel-level supervision to optimize the geometry, significantly improving the reconstruction accuracy of fine-grained human details. To make this approach practical, we further propose a modular data annotation strategy that enriches the training data and enhances the robustness of the model. Overall, PEAR is a preprocessing-free framework that can simultaneously infer EHM-s (SMPLX and scaled-FLAME) parameters at over 100 FPS. Extensive experiments on multiple benchmark datasets demonstrate that our method achieves substantial improvements in pose estimation accuracy compared to previous SMPLX-based approaches. Project page: https://wujh2001.github.io/PEAR

cs.CV↗
arXiv2026

ATOMS-QUARKS survey: Inflow and infall in massive protocluster G318.049+00.086: Evidence of competitive accretion

We present a gas kinematic study of the massive protocluster G318.049+00.086. The protocluster is reported to contain 12 prestellar core candidates and 4 protostellar cores. Filamentary structures are identified using the 1.3 mm dust continuum map, with four of them converge into a dense central region, forming a hub-filament system (HFS). High velocity gradients (10 - 20 km s$^{-1}$ pc$^{-1}$) derived from PV analysis of H$^{13}$CO$^{+}$ emission along three of those filaments are suggestive of mass inflow onto the central hub. A mass inflow rate higher than $10^{3}$ M$_{\odot}$ Myr$^{-1}$ along the filaments is indicating that the central hub is capable of forming massive star(s). Investigation of H$^{13}$CO$^{+}$ and CCH spectral profiles revealed the majority of the cores having the characteristic blue asymmetric line profiles, typical signature of gravitational collapse. The remaining few cores showed red asymmetric profiles, indicative of gas expansion. Also, the derived mass infall rates for the protostellar cores in hub-region is significantly higher in comparison to those located along the filaments. The mass-radius relationship of the cores revealed that the cores with red profiles reside in the massive star formation regime. However, the global velocity gradient along the filaments suggests that these particular cores are losing material to the hub. Our results are supporting a competitive accretion scenario of massive star formation where gas is expected to be funnelled from less gravitationally dominant cores to the cores located at the gravitationally favorable position.

astro-ph.GA↗