arXiv subjects

Dong Li

Publications and source records attributed to Dong Li.

At least 55 records · Page 3Linked to original sources

arXiv2026

Layout-Conditioned Autoregressive Text-to-Image Generation via Structured Masking

Although autoregressive (AR) models have demonstrated remarkable success in image generation, extending these models to layout-conditioned generation remains challenging due to the sparse nature of layout conditions and the risk of feature entanglement. We present \textbf{S}tructured \textbf{M}asking for \textbf{AR}-based \textbf{L}ayout-to-\textbf{I}mage (SMARLI), a novel framework that effectively integrates spatial layout constraints into the AR generation process. To equip AR models with layout control, a structured masking strategy is applied to the attention computation to govern the interaction among the global prompt, layout, and image tokens. This design prevents the misassociation of different regions with their corresponding descriptions while enabling the sufficient injection of layout constraints into the generation process. To alleviate the exposure bias of AR models and further enhance generation quality and layout accuracy, we incorporate a Group Relative Policy Optimization (GRPO) post-training scheme. We adapt it to the next-set-based paradigm and introduce a specifically designed layout reward, which is coordinated with an image quality reward to guide policy optimization in a balanced manner. Experimental results demonstrate that SMARLI seamlessly integrates layout tokens with text and image tokens without compromising generation quality, and the proposed masking strategy and post-training scheme can also be transferred to standard next-token-based AR models. The proposed framework achieves superior layout control while maintaining the structural simplicity and generation efficiency of AR models.

cs.CV↗

arXiv2026

Asymmetric GARCH modelling without moment conditions

Heavy tails and stability are two persistent challenges in modelling financial time series, yet most existing approaches rely on finite-moment assumptions and pay insufficient attention to stability issues. To bridge this gap, we propose an asymmetric GARCH model with standardized non-Gaussian stable innovations (sAGARCH), which accommodates infinite variance and even infinite mean. We establish a comprehensive inference framework for both stationary and explosive cases, proving the strong consistency and asymptotic normality of the maximum likelihood estimator, including the tail index parameter. We also discuss multiple estimators for the asymptotic variance. Additionally, we propose a modified Kolmogorov-type test statistic for diagnostic checking, along with tests for strict stationarity and asymmetry. Through Monte Carlo simulations with heavy-tailed innovations, we provide further insight into the finite-sample performance of the intercept estimator. Empirical applications to stock returns further highlight the usefulness and merits of the proposed sAGARCH model.

stat.ME↗

arXiv2026

SpikeVLA: Vision-Language-Action Models with Spiking Neural Networks

Vision-Language-Action (VLA) models have become a dominant paradigm for embodied intelligence. However, most existing approaches are built on large-scale transformers, resulting in substantial inference latency and energy consumption that limit their practical deployment in low-power, real-time scenarios. We propose SpikeVLA, a spiking VLA architecture for embodied navigation with energy-efficient inference, consisting of three key components. (i) A spiking vision encoder, Spike-V, that replaces dense continuous layers with event-driven spiking layers to reduce the energy consumption of visual representation learning. (ii) A multi-modal spiking large language model, Spike-L, that reformulates cross-modal reasoning with spiking dynamics and token-level event-driven sparsity to further lower computational cost. (iii) A spiking action policy network, Spike-A employs Laplacian-kernel population coding with a multi-layer fully connected SNN, and decodes spiking activities into stable and robust continuous control with energy-efficient inference under low-power constraints. Experiments on navigation and robotic control tasks show that SpikeVLA significantly reduces energy consumption and computational cost while maintaining competitive performance, highlighting its potential for low-power, real-time embodied intelligence.

cs.RO↗

arXiv2026

Unlocking Cryogenic Energy Storage by Constructing Dipole Glass with Unit-cell-level Polar Disorder

Cryogenic energy storage is vital for frontier technologies including deep-space exploration and quantum computing, yet conventional electrochemical energy systems fail below ~230 K due to frozen ion migration. While relaxor-based dielectric capacitors provide high efficiency at room temperature, the intrinsic freezing/growth of polar nanodomains at extended cryogenic regime limits their applications with deteriorated hysteresis losses. Here, we realize superior cryogenic energy-storage performance by designing unit-cell-level disordered dipole-glass state in Pb0.6Sr0.4ZrO3 thin films with composition near antiferroelectric-paraelectric phase boundary. The antiferroelectric-derived dipole-glass introduces enhanced unit-cell-level complexity of dipole interaction that suppresses long-range ferroelectric order. This enables ultralow-hysteresis operation (efficiency > 88%) down to 4 K, delivering record-high energy density (211 J/cm^3) at 9 MV/cm, stability over 10^8 charge/discharge cycles and microsecond-scale charge/discharge capability. This work establishes a dipole-glass paradigm for cryogenic dielectric capacitors, opening a new avenue to highly-efficient energy-storage systems with broad applications in frontier nanoelectronics.

cond-mat.mtrl-sci↗

arXiv2026

Surface-switchable nonreciprocity protected by Fermi arcs in Weyl semimetal TaAs

Weyl semimetals host topologically protected surface states, known as Fermi arcs, which connect bulk Weyl nodes in momentum space. Both bulk Weyl nodes and Fermi arcs are anticipated to be chiral. The chirality of bulk bands has been confirmed through observations of the chiral anomaly and Weyl orbits. In contrast, despite their discovery more than a decade ago, the chiral nature of Fermi arcs has remained unresolved. Here we report Fermi-arc-induced nonlinear transport in the archetypal Weyl semimetal TaAs. Using focused ion beam techniques, we fabricated micro-scale devices that enable simultaneous transport measurements on opposing topological surfaces. While linear transport remains dominated by bulk conduction, nonlinear transport uncovers surface-specific contributions, including an exceptionally large third-order nonreciprocal response that exceeds conventional expectations and highlights the crucial role of the singular arc endpoints. Our findings unambiguously demonstrate the chiral nature of Fermi arcs and establish nonlinear transport as a direct probe of these topological surface states. By revealing a surface-switchable, room-temperature nonlinear response that is topologically protected, this work introduces a new functionality in Weyl semimetals. Given the abundance of natural materials predicted to host topological semimetal states, these results open opportunities for exploring nonlinear transport phenomena and device concepts across a broad class of systems.

cond-mat.mes-hall↗

arXiv2026

The 1st PortraitCraft Challenge: A CVPR 2026 Workshop Competition on Portrait Composition Understanding and Generation

This paper presents an overview of the inaugural PortraitCraft Challenge, held as one of the official competitions at CVPR 2026. The challenge focuses on portrait composition understanding and generation, aiming to advance AI research in portrait aesthetics analysis and controllable image synthesis. Unlike existing datasets and tasks that primarily focus on global aesthetic scoring, PortraitCraft introduces a unified evaluation framework comprising two complementary tracks. Track 1 requires models to perform structured portrait composition understanding, and Track 2 requires models to generate portrait images from structured composition descriptions under explicit compositional constraints. To support the challenge, we constructed and publicly released a large-scale portrait composition dataset consisting of approximately 50,000 curated real portrait images, providing multi-level supervision. This report describes the challenge setup, evaluation protocols, dataset composition, and final results, along with an analysis of the technical characteristics of the submitted solutions. The PortraitCraft Challenge provides a standardized and reproducible platform for research on portrait composition understanding and generation, and is expected to foster further progress in the fields of portrait aesthetics and controllable image generation.

cs.CV↗

arXiv2026

Physics-Informed Neural Network Modeling of Biodegradable Contaminant Transport through GCL/SL Composite Liners

This study develops a two-domain physics-informed neural network framework for contaminant transport through a GCL/SL composite liner system, in which the thin GCL layer is treated using a steady-state advection-dispersion-biodegradation formulation and the underlying soil liner is modeled as a transient transport domain. Two formulations are evaluated against analytical and finite-element reference solutions under different leachate-head conditions: a standard PINN with soft constraint enforcement (Std-PINN) and a hard-constrained PINN (H-PINN), in which selected boundary and initial conditions are embedded directly into the trial solutions. The Std-PINN captures the overall breakthrough behavior but shows larger errors during the early transport stage, particularly under higher leachate heads where advective transport becomes more pronounced. The H-PINN reduces the optimization burden associated with penalty-based constraint enforcement and provides more accurate and stable concentration predictions, lowering the MAE from approximately 0.058-0.067 for the Std-PINN to about 0.011-0.023 for the H-PINN, while reducing the MRE from approximately 9.10%-19.16% to about 2.08%-3.14%. Parametric analyses confirm that the H-PINN with the tanh activation function and an optimized network structure provides the best predictive accuracy. The H-PINN is further extended to inverse modeling for identifying the SL degradation half-life from limited concentration observations, showing reliable convergence toward prescribed values and acceptable robustness under low-to-moderate observation noise.

cs.LG↗

arXiv2026

PRISM: Synergizing Vision Foundation Models via Self-organized Expert Specialization

Unifying the complementary strengths of diverse Vision Foundation Models (VFMs) into a single efficient model is highly desirable but challenged by the negative transfer inherent in monolithic distillation. To address these feature conflicts, we introduce \textbf{PRISM}, a novel dual-stream Mixture-of-Experts (MoE) framework that synergizes VFMs via modular specialization. We propose a two-stage paradigm: (1) expertise deconstruction, where a teacher-conditional router guides experts to specialize in distinct representational subspaces to mitigate interference, followed by (2) dynamic recomposition, where the router learns to assemble these experts into tailored computational pathways for downstream tasks. Experiments on PASCAL-Context and NYUD-v2 show that \textbf{PRISM} establishes a new state of the art, validating that sparse, emergent specialization is a scalable approach for integrating diverse visual knowledge.

cs.CV↗

arXiv2026

Unveiling the Entropy Dynamics of Chain-of-Thought Reasoning

This paper investigates the entropy dynamics of Chain-of-Thought (CoT) and uncovers a consistent two-phase structure: an Uncertainty Region of exploration transitioning sharply to a Confidence Region of convergence. We demonstrate that the Confidence Region possesses two critical properties: 1) High Reliability -- answers in the confidence region become highly accurate and stable, and 2) High Redundancy -- models generate unnecessary tokens long after reaching the correct answer. These properties unlock more efficient and reliable inference strategies: 1) Early Exit leverages reliability and redundancy to terminate computation safely when returns diminish, and 2)Test-Time Scaling uses the Confidence Region signal to prioritize converged trajectories. To operationalize these insights, we formulate Confidence Region detection as a sequential change-point detection problem, being the first to apply classical change-point methods to monitor CoT reasoning. Using the Cumulative Sum (CUSUM) algorithm, a statistically optimal change-point detector, we develop a training-free framework for real-time inference control. Experiments show our approach establishes a superior Pareto-frontier for early exit. CUSUM achieves 63.06% accuracy with 11.1% token reduction, outperforming DEER and Dynasor by 3.28% and 4.36% in accuracy respectively. For test-time scaling, CUSUM-weighted voting consistently outperforms self-consistency.

cs.CL↗

arXiv2026

CL-DMDF:Dynamic Multimodal Data Fusion Model Based on Contrastive Learning

Multimodal data fusion involves integrating and analyzing information from multiple modalities to uncover latent correlations and complementary patterns, thereby enhancing data processing and decision-making. While existing methods for structured multimodal inputs are typically designed around specific tasks and assume fully observed modalities, real-world applications often suffer from uncertain or missing modality inputs due to various factors. Some traditional models overly emphasize local interactions within missing modalities, neglecting the global complementary cues embedded in multimodal representations. To overcome these limitations, we propose a Dynamic Multimodal Data Fusion model based on Contrastive Learning (CL-DMDF). CL-DMDF introduces a novel attention mechanism that operates across both feature and modality dimensions to compute reliable attention scores, effectively reflecting importance at each level. The CL-DMDF further incorporates an entity-centroid contrastive learning module that constructs centroid-based positive samples from entity features to enhance discriminative learning. Additionally, an adaptive fusion module is employed to improve the efficiency and accuracy of dynamic fusion strategies. Extensive experiments conducted on three datasets demonstrate the effectiveness of the CL-DMDF across diverse multimodal fusion tasks.

cs.LG↗

arXiv2026

Efficient Diffusion LLMs via Temporal-Spatial Parallel Decoding and Confidence Extrapolation

Diffusion-based large language models (dLLMs) support parallel text generation via iterative denoising, yet inference remains latency-heavy because many steps are spent on redundant refinement and repeated remasking of tokens whose final values are already determined. Prior acceleration methods mainly depend on step-local confidence heuristics or fixed schedules, which are sensitive to prompt and task variation and ignore strong positional effects within a sequence. We cast diffusion decoding as a dynamic control problem and show that token-wise denoising trajectories provide the key signal for reliable control. We propose a trace-aware decoding framework with two components. First, Temporal-Spatial Parallel Decoding (TSPD) uses a lightweight temporalspatial controller that consumes per-token trajectory features, including confidence, entropy, and momentum, together with token position, to decide when a token has converged and can be safely fixed. Second, we introduce Confidence Extrapolation (CE), a training-free state-space module that forecasts future logit trends with uncertainty to support proactive decisions, including safe look-ahead and targeted stabilization when trajectories are oscillatory or underconfident. Together, TSPD and CE reduce unnecessary denoising iterations while preserving output quality, and they compose cleanly with system optimizations such as KV caching.

cs.CL↗

arXiv2026

Athena: Enhancing Multimodal Reasoning with Data-efficient Process Reward Models

We present Athena-PRM, a multimodal process reward model (PRM) designed to evaluate the reward score for each step in solving complex reasoning problems. Developing high-performance PRMs typically demands significant time and financial investment, primarily due to the necessity for step-level annotations of reasoning steps. Conventional automated labeling methods, such as Monte Carlo estimation, often produce noisy labels and incur substantial computational costs. To efficiently generate high-quality process-labeled data, we propose leveraging prediction consistency between weak and strong completers as a criterion for identifying reliable process labels. Remarkably, Athena-PRM demonstrates outstanding effectiveness across various scenarios and benchmarks with just 5,000 samples. Furthermore, we also develop two effective strategies to improve the performance of PRMs: ORM initialization and up-sampling for negative data. We validate our approach in three specific scenarios: verification for test time scaling, direct evaluation of reasoning step correctness, and reward ranked fine-tuning. Our Athena-PRM consistently achieves superior performance across multiple benchmarks and scenarios. Notably, when using Qwen2.5-VL-7B as the policy model, Athena-PRM enhances performance by 10.2 points on WeMath and 7.1 points on MathVista for test time scaling. Furthermore, Athena-PRM sets the state-of-the-art (SoTA) results in VisualProcessBench and outperforms the previous SoTA by 3.9 F1-score, showcasing its robust capability to accurately assess the correctness of the reasoning step. Additionally, utilizing Athena-PRM as the reward model, we develop Athena-7B with reward ranked fine-tuning and outperforms baseline with a significant margin on five benchmarks.

cs.LG↗

arXiv2026

Ratio-Variance Regularized Policy Optimization

Standard on-policy reinforcement learning relies on heuristic clipping to enforce trust regions, but this mechanism imposes a severe cost by indiscriminately truncating high-return yet high-divergence updates. We demonstrate that explicitly constraining the policy ratio variance provides a principled local approximation to trust-region constraints, eliminating the need for binary hard clipping. By acting as a distributional ``soft brake'', this approach preserves critical gradient signals from novel discoveries while naturally down-weighting and enabling the reuse of stale, off-policy data. We introduce ${\bf R}^2{\bf VPO}$ (Ratio-Variance Regularized Policy Optimization), which implements this constraint via a primal-dual optimization framework. Extensive evaluations across $7$ LLM scales, spanning both fast and slow reasoning paradigms, and $10$ robotic control tasks demonstrate the generality of the proposed approach. R$^2$VPO achieves substantial performance gains on mathematical reasoning benchmarks, with particularly pronounced improvements on smaller models, while significantly improving sample efficiency. Furthermore, it consistently outperforms PPO baselines in continuous control domains, particularly in sparse-reward and dynamic environments. Together, these findings establish ratio-variance regularization as a principled foundation for stable and data-efficient policy optimization.

cs.LG↗

arXiv2026

DSRDM: Digital Signal Recovery Diffusion Model for Semantic Communications

Diffusion model (DM) has recently appeared as a promising type of generative model for AI-generated content, which has been widely used for image reconstruction, generation, and channel denoising in semantic communication (SemCom) due to its strong generation capacity. However, most of existing works regarding SemCom remain confined to the image or text transmission, and neglect the commonly adopted digital signals in wireless systems. In this letter, in order to address this gap, we propose and investigate a digital signal recovery diffusion model (DSRDM) for SemCom. Specifically, DSRDM encodes digital signals by gradually adding Gaussian signals to images in the forward diffusion process of DM. After the encoded Gaussian signals embedded in the carrier image are sent to the receiver, it recovers the digital signals by predicting the added Gaussian signals iteratively in the reverse diffusion process. Moreover, to reduce the computation complexity of DSRDM, a signal adding approach is designed to avoid the retraining latency. In particular, we use the latent representation of images instead of themselves as the carrier for digital signals in DSRDM to reduce the inference latency.

eess.SP↗

arXiv2026

Dynamic Dual-Granularity Skill Bank for Agentic RL

Agentic RL can benefit substantially from reusable experience, yet existing skill-based methods mainly extract trajectory-level guidance and often lack principled mechanisms for maintaining an evolving skill memory. We propose D2Skill, a dynamic dual-granularity skill bank for agentic RL that organizes reusable experience into task skills for high-level guidance and step skills for fine-grained decision support and error correction. D2Skill jointly trains the policy and skill bank through paired baseline and skill-injected rollouts under the same policy, using their performance gap to derive hindsight utility signals for both skill updating and policy optimization. Built entirely from training-time experience, the skill bank is continuously expanded through reflection and maintained with utility-aware retrieval and pruning. Experiments on ALFWorld, WebShop, and Search-Augmented QA tasks show that D2Skill substantially improves performance over skill-free baselines across models of different scales. Further ablations and analyses show that both dual-granularity skill modeling and dynamic skill maintenance are critical to these gains, while the learned skills exhibit higher utility, transfer across evaluation settings, and introduce only modest training overhead.

cs.AI↗

arXiv2026

Beyond the Target: From Imitation to Collaboration in Speculative Decoding

Speculative decoding (SPD) accelerates large language model (LLM) inference by letting a smaller draft model propose multiple future tokens that are verified in parallel by a larger target model. The dominant SPD paradigm treats the target model as the sole reliable teacher, accepting a draft token only when it exactly matches the target prediction. This design implicitly assumes that the target is always the better choice at every position. In practice, this assumption does not hold. Although the draft is the weaker model overall, it is not uniformly inferior at the token level. In a meaningful fraction of cases where draft and target disagree, the draft's choice is the one that leads to the correct final answer. Inspired by this, we introduce \textbf{Collaborative Speculative Decoding (CoSpec)}, a generalization of SPD that no longer treats the target model as the sole token-level authority. CoSpec trains an arbitration policy via reinforcement learning to decide whether to accept tokens from the draft or target model, selectively accepting draft tokens at mismatches when doing so is likely to yield a correct final answer. Experimental results show that CoSpec maintains substantial speedups while surpassing target-only performance. By shifting the emphasis from imitation to collaboration, CoSpec suggests a new perspective on speculative decoding.

cs.CL↗

arXiv2026

Coupled Hierarchical Search over Topology and Execution for Agentic Workflow Synthesis

Although structured workflows empower Large Language Models (LLMs) to tackle complex problems, automating their creation is severely hindered by a vast combinatorial search space, frequently resulting in inflexible and resource-heavy offline training dependencies. To address this, we conceptualize workflow generation as an intertwined topology-and-execution search paradigm, where the broader topological layer dictates subtask boundaries and lower-level execution outcomes actively reshape the topology itself. Building on this foundation, we introduce HierFlow, a training-free, test-time hierarchical search architecture that automates agentic workflow design by merging feedback-guided topology adjustments with a fast, MCTS-inspired tree search for sub-workflow optimization. HierFlow maximizes efficiency through an intelligent gating module that selectively triggers execution-level searches based on contextual necessity, a mechanism we further support with an in-depth analysis detailing how varying degrees of cross-task coupling impact the effectiveness of hierarchical splitting. Comprehensive testing across question answering, mathematical reasoning, and code generation benchmarks confirms that HierFlow consistently outperforms strong baselines, delivering an optimal balance of high-quality results and computational efficiency without any additional training overhead.

cs.AI↗

arXiv2026

MIITA: Memory-Induced Inference-Time Adaptation for Continual Learning with Small Language Models

Continual learning (CL) is essential for small language models (SLMs) to adapt to evolving real-world needs in resource-constrained deployments. However, directly updating their limited parameter space causes catastrophic forgetting. While memory-based methods naturally address this by decoupling knowledge retention from parameters, existing approaches designed for large language models (LLMs) rely on abundant storage and strong in-context reasoning that SLMs lack. To address these challenges, we propose MIITA, a Memory-Induced Inference-Time Adaptation framework for supervised CL under constrained storage. MIITA stores supervised experiences as compact correction-direction prototypes with semantic anchors, and retrieves them at inference time using semantic and uncertainty-based cues. The retrieved directions are applied through gated temporary hidden-state adaptation, enabling non-destructive reuse of past supervision without backbone updates, prompt extensions, or test-time backpropagation. A local theoretical analysis links this design to first-order loss reduction, uncertainty-guided retrieval, and directional coverage for retaining old-stage knowledge. Extensive experiments across diverse supervised CL settings show that MIITA consistently improves final performance and mitigates forgetting under fixed memory budgets.

cs.AI↗