arXiv subjects

Wei Li

Publications and source records attributed to Wei Li.

At least 271 records · Page 15Linked to original sources

arXiv2026

Branch, or Layer? Zeroth-Order Optimization for Continual Learning of Vision-Language Models

Vision-Language Continual Learning (VLCL) has attracted significant research attention for its robust capabilities, and the adoption of Parameter-Efficient Fine-Tuning (PEFT) strategies is enabling these models to achieve competitive performance with substantially reduced resource consumption. However, dominated First-Order (FO) optimization is prone to trap models in suboptimal local minima, especially in limited exploration subspace within PEFT. To overcome this challenge, this paper pioneers a systematic exploration of adopting Zeroth-Order (ZO) optimization for PEFT-based VLCL. We first identify the incompatibility of naive full-ZO adoption in VLCL due to optimization process instability. We then investigate the application of ZO optimization from a modality branch-wise to a fine-grained layer-wise across various training units to identify an optimal strategy. Besides, a key theoretical insight reveals that vision modality exhibit higher variance than language counterparts in VLCL during the ZO optimization process, and we propose a modality-aware ZO strategy, which adopts gradient sign normalization in ZO and constrains vision modality perturbation to further improve performance. Benefiting from the adoption of ZO optimization, PEFT-based VLCL fulfills better ability to escape local minima during the optimization process, extensive experiments on four benchmarks demonstrate that our method achieves state-of-the-art results.

cs.CV↗

arXiv2026

CliCARE: Grounding Large Language Models in Clinical Guidelines for Decision Support over Longitudinal Cancer Electronic Health Records

Large Language Models (LLMs) hold significant promise for improving clinical decision support and reducing physician burnout by synthesizing complex, longitudinal cancer Electronic Health Records (EHRs). However, their implementation in this critical field faces three primary challenges: the inability to effectively process the extensive length and fragmented nature of patient records for accurate temporal analysis; a heightened risk of clinical hallucination, as conventional grounding techniques such as Retrieval-Augmented Generation (RAG) do not adequately incorporate process-oriented clinical guidelines; and unreliable evaluation metrics that hinder the validation of AI systems in oncology. To address these issues, we propose CliCARE, a framework for Grounding Large Language Models in Clinical Guidelines for Decision Support over Longitudinal Cancer Electronic Health Records. The framework operates by transforming unstructured, longitudinal EHRs into patient-specific Temporal Knowledge Graphs (TKGs) to capture long-range dependencies, and then grounding the decision support process by aligning these real-world patient trajectories with a normative guideline knowledge graph. This approach provides oncologists with evidence-grounded decision support by generating a high-fidelity clinical summary and an actionable recommendation. We validated our framework using large-scale, longitudinal data from a private Chinese cancer dataset and the public English MIMIC-IV dataset. In these settings, CliCARE significantly outperforms baselines, including leading long-context LLMs and Knowledge Graph-enhanced RAG methods. The clinical validity of our results is supported by a robust evaluation protocol, which demonstrates a high correlation with assessments made by oncologists.

cs.CL↗

arXiv2026

Learning from Mistakes: Negative Reasoning Samples Enhance Out-of-Domain Generalization

Supervised fine-tuning (SFT) on chain-of-thought (CoT) trajectories demonstrations is a common approach for enabling reasoning in large language models. Standard practices typically only retain trajectories with correct final answers (positives) while ignoring the rest (negatives). We argue that this paradigm discards substantial supervision and exacerbates overfitting, limiting out-of-domain (OOD) generalization. Specifically, we surprisingly find that incorporating negative trajectories into SFT yields substantial OOD generalization gains over positive-only training, as these trajectories often retain valid intermediate reasoning despite incorrect final answers. To understand this effect in depth, we systematically analyze data, training dynamics, and inference behavior, identifying 22 recurring patterns in negative chains that serve a dual role: they moderate loss descent to mitigate overfitting during training and boost policy entropy by 35.67% during inference to facilitate exploration. Motivated by these observations, we further propose Gain-based LOss Weighting (GLOW), an adaptive, sample-aware scheme that exploits such distinctive training dynamics by rescaling per-sample loss based on inter-epoch progress. Empirically, GLOW efficiently leverages unfiltered trajectories, yielding a 5.51% OOD gain over positive-only SFT on Qwen2.5-7B and boosting MMLU from 72.82% to 76.47% as an RL initialization.

cs.CL↗

arXiv2026

EntroLnn: Entropy-Guided Liquid Neural Networks for Operando Refinement of Battery Capacity Fade Trajectories

Battery capacity degradation prediction has long been a central topic in battery health analytics, and most studies focus on state of health (SoH) estimation and end of life (EoL) prediction. This study extends the scope to online refinement of the entire capacity fade trajectory (CFT) through EntroLnn, a framework based on entropy-guided transformable liquid neural networks (LNNs). EntroLnn treats CFT refinement as an integrated process rather than two independent tasks for pointwise SoH and EoL. We introduce entropy-based features derived from online temperature fields, applied for the first time in battery analytics, and combine them with customized LNNs that model temporal battery dynamics effectively. The framework enhances both static and dynamic adaptability of LNNs and achieves robust and generalizable CFT refinement across different batteries and operating conditions. The approach provides a high fidelity battery health model with lightweight computation, achieving mean absolute errors of only 0.004577 for CFT and 18 cycles for EoL prediction. This work establishes a foundation for entropy-informed learning in battery analytics and enables self-adaptive, lightweight, and interpretable battery health prediction in practical battery management systems.

cs.LG↗

arXiv2026

Dominant Kitaev Interaction and Field-induced Quantum Disordered Phase in the Cobaltate Na$_2$Co$_2$TeO$_6$

The identification of quantum spin liquid phases in Kitaev candidate materials remains a major experimental challenge. Since most Kitaev candidates develop antiferromagnetic (AFM) order at low temperatures, currently there are great interest on the field-induced magnetic disordered phase in these compounds, that are distinct from (partially) polarized states. Recently, a cobaltate Na$_2$Co$_2$TeO$_6$ has emerged as a promising Kitaev candidate with high-spin $t^{5}_{2g}e^2_g$ configuration and spin-orbit entangled $J_{\rm eff} = 1/2$ honeycomb lattice system. There are intensive studies on field-induced magnetic states and phase transitions under in-plane magnetic fields. In this study, we propose an intermediate disordered phase induced by an out-of-plane field along the $c$-axis, through high-field magnetization and magnetocaloric effect measurements. To explain the high-field behavior of Na$_2$Co$_2$TeO$_6$, we develop an effective $K$-$J$-$Γ$-$Γ^{\prime}$ spin model featuring a dominant AFM Kitaev interaction. This framework uncovers an intermediate quantum spin liquid phase, establishing the material as a unique platform for exploring Kitaev physics and field-induced quantum-disordered states.

cond-mat.str-el↗

arXiv2026

A new partial differential nonlinear system containing quasivariational and parabolic variational inequalities and its application

We study a new nonlinear system which contains a partial differential equation, a quasivariational inequality and a parabolic variational inequality in Banach spaces. We obtain the unique solvability of the coupled system under moderate conditions by using the Banach's fixed point theorem. We employ the main results to investigate a viscoelastic frictional contact problem with long-memory effects, wear processes, and damage phenomenon.

math.AP↗

arXiv2025

Controllable Human-centric Keyframe Interpolation with Generative Prior

Existing interpolation methods use pre-trained video diffusion priors to generate intermediate frames between sparsely sampled keyframes. In the absence of 3D geometric guidance, these methods struggle to produce plausible results for complex, articulated human motions and offer limited control over the synthesized dynamics. In this paper, we introduce PoseFuse3D Keyframe Interpolator (PoseFuse3D-KI), a novel framework that integrates 3D human guidance signals into the diffusion process for Controllable Human-centric Keyframe Interpolation (CHKI). To provide rich spatial and structural cues for interpolation, our PoseFuse3D, a 3D-informed control model, features a novel SMPL-X encoder that transforms 3D geometry and shape into the 2D latent conditioning space, alongside a fusion network that integrates these 3D cues with 2D pose embeddings. For evaluation, we build CHKI-Video, a new dataset annotated with both 2D poses and 3D SMPL-X parameters. We show that PoseFuse3D-KI consistently outperforms state-of-the-art baselines on CHKI-Video, achieving a 9% improvement in PSNR and a 38% reduction in LPIPS. Comprehensive ablations demonstrate that our PoseFuse3D model improves interpolation fidelity.

cs.CV↗

arXiv2025

From Sequential to Spatial: Reordering Autoregression for Efficient Visual Generation

Inspired by the remarkable success of autoregressive models in language modeling, this paradigm has been widely adopted in visual generation. However, the sequential token-by-token decoding mechanism inherent in traditional autoregressive models leads to low inference efficiency.In this paper, we propose RadAR, an efficient and parallelizable framework designed to accelerate autoregressive visual generation while preserving its representational capacity. Our approach is motivated by the observation that visual tokens exhibit strong local dependencies and spatial correlations with their neighbors--a property not fully exploited in standard raster-scan decoding orders. Specifically, we organize the generation process around a radial topology: an initial token is selected as the starting point, and all other tokens are systematically grouped into multiple concentric rings according to their spatial distances from this center. Generation then proceeds in a ring-wise manner, from inner to outer regions, enabling the parallel prediction of all tokens within the same ring. This design not only preserves the structural locality and spatial coherence of visual scenes but also substantially increases parallelization. Furthermore, to address the risk of inconsistent predictions arising from simultaneous token generation with limited context, we introduce a nested attention mechanism. This mechanism dynamically refines implausible outputs during the forward pass, thereby mitigating error accumulation and preventing model collapse. By integrating radial parallel prediction with dynamic output correction, RadAR significantly improves generation efficiency.

cs.CV↗

arXiv2025

TeleWorld: Towards Dynamic Multimodal Synthesis with a 4D World Model

World models aim to endow AI systems with the ability to represent, generate, and interact with dynamic environments in a coherent and temporally consistent manner. While recent video generation models have demonstrated impressive visual quality, they remain limited in real-time interaction, long-horizon consistency, and persistent memory of dynamic scenes, hindering their evolution into practical world models. In this report, we present TeleWorld, a real-time multimodal 4D world modeling framework that unifies video generation, dynamic scene reconstruction, and long-term world memory within a closed-loop system. TeleWorld introduces a novel generation-reconstruction-guidance paradigm, where generated video streams are continuously reconstructed into a dynamic 4D spatio-temporal representation, which in turn guides subsequent generation to maintain spatial, temporal, and physical consistency. To support long-horizon generation with low latency, we employ an autoregressive diffusion-based video model enhanced with Macro-from-Micro Planning (MMPL)--a hierarchical planning method that reduces error accumulation from frame-level to segment-level-alongside efficient Distribution Matching Distillation (DMD), enabling real-time synthesis under practical computational budgets. Our approach achieves seamless integration of dynamic object modeling and static scene representation within a unified 4D framework, advancing world models toward practical, interactive, and computationally accessible systems. Extensive experiments demonstrate that TeleWorld achieves strong performance in both static and dynamic world understanding, long-term consistency, and real-time generation efficiency, positioning it as a practical step toward interactive, memory-enabled world models for multimodal generation and embodied intelligence.

cs.CV↗

arXiv2025

Mathematical Theory for Photonic Hall Effect in Honeycomb Photonic Crystals

In this work, we develop a mathematical theory for the photonic Hall effect and prove the existence of guided electromagnetic waves at the interface of two honeycomb photonic crystals. The guided wave resembles the edge states in electronic systems: it is induced by the topological Hall effect, and the wave propagates along the interface but not in the bulk media. Starting from a symmetric honeycomb photonic crystal that attains Dirac points at the high-symmetry points of the Brillouin zone, $K$ and $K'$, we introduce two classes of perturbations for the periodic medium. The perturbations lift the Dirac degeneracy, forming a spectral band valley at the points $K$ and $K'$ with well-defined topological phase that depends on the sign of the perturbation parameters. By employing the layer potential techniques and spectral analysis, we investigate the existence of guided wave along an interface when two honeycomb photonic crystals are glued together. In particular, we elucidate the relationship between the existence of the interface mode and the nature of perturbations imposed on the two periodic media separated by the interface.

physics.optics↗

arXiv2025

Adaptive Trajectory Bundle Method for Roll-to-Roll Manufacturing Systems

Roll-to-roll (R2R) manufacturing requires precise tension and velocity control under operational constraints. Model predictive control demands gradient computation, while sampling-based methods like MPPI struggle with hard constraint satisfaction. This paper presents an adaptive trajectory bundle method that achieves rigorous constraint handling through derivative-free sequential convex programming. The approach approximates nonlinear dynamics and costs via interpolated sample bundles, replacing Taylor-series linearization with function-value interpolation. Adaptive trust region and penalty mechanisms automatically adjust based on constraint violation metrics, eliminating manual tuning. We establish convergence guarantees proving finite-time feasibility and convergence to stationary points of the constrained problem. Simulations on a six-zone R2R system demonstrate that the adaptive method achieves 4.3\% lower tension RMSE than gradient-based MPC and 11.1\% improvement over baseline TBM in velocity transients, with superior constraint satisfaction compared to MPPI variants. Experimental validation on an R2R dry transfer system confirms faster settling and reduced overshoot relative to LQR and non-adaptive TBM.

eess.SY↗

arXiv2025

ESearch-R1: Learning Cost-Aware MLLM Agents for Interactive Embodied Search via Reinforcement Learning

Multimodal Large Language Models (MLLMs) have empowered embodied agents with remarkable capabilities in planning and reasoning. However, when facing ambiguous natural language instructions (e.g., "fetch the tool" in a cluttered room), current agents often fail to balance the high cost of physical exploration against the cognitive cost of human interaction. They typically treat disambiguation as a passive perception problem, lacking the strategic reasoning to minimize total task execution costs. To bridge this gap, we propose ESearch-R1, a cost-aware embodied reasoning framework that unifies interactive dialogue (Ask), episodic memory retrieval (GetMemory), and physical navigation (Navigate) into a single decision process. We introduce HC-GRPO (Heterogeneous Cost-Aware Group Relative Policy Optimization). Unlike traditional PPO which relies on a separate value critic, HC-GRPO optimizes the MLLM by sampling groups of reasoning trajectories and reinforcing those that achieve the optimal trade-off between information gain and heterogeneous costs (e.g., navigate time, and human attention). Extensive experiments in AI2-THOR demonstrate that ESearch-R1 significantly outperforms standard ReAct-based agents. It improves task success rates while reducing total operational costs by approximately 50\%, validating the effectiveness of GRPO in aligning MLLM agents with physical world constraints.

cs.AI↗

arXiv2025

Lattice-decoupled rotatable stripe-like charge order within the strange metal phase of 2M-WS2

In quantum materials, charge orders typically stabilize in specific crystallographic orientations, though their formation mechanisms may vary. Here, using low-temperature scanning tunneling microscopy (STM), we discover a lattice-decoupled rotatable stripe-like charge order coexisting with superconductivity in 2M-WS2. The charge order manifests five distinct orientations across different sample regions, yet maintains an identical wavelength. This directional decoupling from host lattice challenges existing paradigms. First-principles calculations of phonon spectra and nesting function fail to explain the ordering mechanism. Intriguingly, the transition temperature of the charge orders exhibits spatial variations (21-46 K), coinciding with the temperature range of the recently reported strange metal phase in this material. This correlation suggests that the interplay between strong electronic correlations and electron-phonon coupling must be critically evaluated to elucidate the emergence of this unconventional charge order.

cond-mat.supr-con↗

arXiv2025

Region-Constraint In-Context Generation for Instructional Video Editing

The In-context generation paradigm recently has demonstrated strong power in instructional image editing with both data efficiency and synthesis quality. Nevertheless, shaping such in-context learning for instruction-based video editing is not trivial. Without specifying editing regions, the results can suffer from the problem of inaccurate editing regions and the token interference between editing and non-editing areas during denoising. To address these, we present ReCo, a new instructional video editing paradigm that novelly delves into constraint modeling between editing and non-editing regions during in-context generation. Technically, ReCo width-wise concatenates source and target video for joint denoising. To calibrate video diffusion learning, ReCo capitalizes on two regularization terms, i.e., latent and attention regularization, conducting on one-step backward denoised latents and attention maps, respectively. The former increases the latent discrepancy of the editing region between source and target videos while reducing that of non-editing areas, emphasizing the modification on editing area and alleviating outside unexpected content generation. The latter suppresses the attention of tokens in the editing region to the tokens in counterpart of the source video, thereby mitigating their interference during novel object generation in target video. Furthermore, we propose a large-scale, high-quality video editing dataset, i.e., ReCo-Data, comprising 500K instruction-video pairs to benefit model training. Extensive experiments conducted on four major instruction-based video editing tasks demonstrate the superiority of our proposal.

cs.CV↗

arXiv2025

Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows

Despite advances in scientific AI, a coherent framework for Scientific General Intelligence (SGI)-the ability to autonomously conceive, investigate, and reason across scientific domains-remains lacking. We present an operational SGI definition grounded in the Practical Inquiry Model (PIM: Deliberation, Conception, Action, Perception) and operationalize it via four scientist-aligned tasks: deep research, idea generation, dry/wet experiments, and experimental reasoning. SGI-Bench comprises over 1,000 expert-curated, cross-disciplinary samples inspired by Science's 125 Big Questions, enabling systematic evaluation of state-of-the-art LLMs. Results reveal gaps: low exact match (10--20%) in deep research despite step-level alignment; ideas lacking feasibility and detail; high code executability but low execution result accuracy in dry experiments; low sequence fidelity in wet protocols; and persistent multimodal comparative-reasoning challenges. We further introduce Test-Time Reinforcement Learning (TTRL), which optimizes retrieval-augmented novelty rewards at inference, enhancing hypothesis novelty without reference answer. Together, our PIM-grounded definition, workflow-centric benchmark, and empirical insights establish a foundation for AI systems that genuinely participate in scientific discovery.

cs.AI↗

arXiv2025

Magnetically Mediated Cross-Layer Pairing in Pressurized Trilayer Nickelate La$_4$Ni$_3$O$_{10}$

The recently discovered trilayer nickelate superconductor La$_4$Ni$_3$O$_{10}$ under pressure has emerged as a promising platform for exploring unconventional superconductivity. However, the pairing mechanism remains a subject of active investigations. With large-scale density matrix renormalization group calculations on a realistic two-orbital trilayer Hubbard model, we elucidate the superconducting (SC) mechanism in this system. Our results reveal distinct magnetic correlations in the two different orbitals: while the $d_{z^2}$ orbital exhibits both interlayer and cross-layer antiferromagnetic (AFM) correlations, the $d_{x^2-y^2}$ orbital shows exclusively cross-layer AFM correlations, rendering a quasi-long-range SC order in the latter. We demonstrate that the Hund's rule coupling is essential for forming the SC order, and discuss the effects of kinetic AFM correlation and Hubbard repulsive $U$. Our findings motivate a further simplification of the trilayer Hubbard to an effective bilayer mixed-dimensional Hubbard model, providing a unified framework for understanding interlayer SC in both trilayer and bilayer nickelates.

cond-mat.str-el↗

arXiv2025

Optimizing the Adversarial Perturbation with a Momentum-based Adaptive Matrix

Generating adversarial examples (AEs) can be formulated as an optimization problem. Among various optimization-based attacks, the gradient-based PGD and the momentum-based MI-FGSM have garnered considerable interest. However, all these attacks use the sign function to scale their perturbations, which raises several theoretical concerns from the point of view of optimization. In this paper, we first reveal that PGD is actually a specific reformulation of the projected gradient method using only the current gradient to determine its step-size. Further, we show that when we utilize a conventional adaptive matrix with the accumulated gradients to scale the perturbation, PGD becomes AdaGrad. Motivated by this analysis, we present a novel momentum-based attack AdaMI, in which the perturbation is optimized with an interesting momentum-based adaptive matrix. AdaMI is proved to attain optimal convergence for convex problems, indicating that it addresses the non-convergence issue of MI-FGSM, thereby ensuring stability of the optimization process. The experiments demonstrate that the proposed momentum-based adaptive matrix can serve as a general and effective technique to boost adversarial transferability over the state-of-the-art methods across different networks while maintaining better stability and imperceptibility.

cs.LG↗

arXiv2025

Light-X: Generative 4D Video Rendering with Camera and Illumination Control

Recent advances in illumination control extend image-based methods to video, yet still facing a trade-off between lighting fidelity and temporal consistency. Moving beyond relighting, a key step toward generative modeling of real-world scenes is the joint control of camera trajectory and illumination, since visual dynamics are inherently shaped by both geometry and lighting. To this end, we present Light-X, a video generation framework that enables controllable rendering from monocular videos with both viewpoint and illumination control. 1) We propose a disentangled design that decouples geometry and lighting signals: geometry and motion are captured via dynamic point clouds projected along user-defined camera trajectories, while illumination cues are provided by a relit frame consistently projected into the same geometry. These explicit, fine-grained cues enable effective disentanglement and guide high-quality illumination. 2) To address the lack of paired multi-view and multi-illumination videos, we introduce Light-Syn, a degradation-based pipeline with inverse-mapping that synthesizes training pairs from in-the-wild monocular footage. This strategy yields a dataset covering static, dynamic, and AI-generated scenes, ensuring robust training. Extensive experiments show that Light-X outperforms baseline methods in joint camera-illumination control and surpasses prior video relighting methods under both text- and background-conditioned settings.

cs.CV↗