arXiv subjects

Dong Li

Publications and source records attributed to Dong Li.

At least 109 records · Page 6Linked to original sources

arXiv2026

On the gap property of a linearized NLS operator

We consider a pair of linear operators corresponding to the linearization around the ground state soliton of the cubic nonlinear Schrödinger equation in dimension three. We introduce a new comparison-based approach and rigorously prove that the interval $(0, 1]$ does not contain any eigenvalues of these operators. Furthermore we show the absence of resonances at the bottom of the essential spectrum. All the obtained results are for the fully non-radial case. The method can be adapted to many other spectral problems.

math.AP↗

arXiv2026

JEPA-VLA: Video Predictive Embedding is Needed for VLA Models

Recent vision-language-action (VLA) models built upon pretrained vision-language models (VLMs) have achieved significant improvements in robotic manipulation. However, current VLAs still suffer from low sample efficiency and limited generalization. This paper argues that these limitations are closely tied to an overlooked component, pretrained visual representation, which offers insufficient knowledge on both aspects of environment understanding and policy prior. Through an in-depth analysis, we find that commonly used visual representations in VLAs, whether pretrained via language-image contrastive learning or image-based self-supervised learning, remain inadequate at capturing crucial, task-relevant environment information and at inducing effective policy priors, i.e., anticipatory knowledge of how the environment evolves under successful task execution. In contrast, we discover that predictive embeddings pretrained on videos, in particular V-JEPA 2, are adept at flexibly discarding unpredictable environment factors and encoding task-relevant temporal dynamics, thereby effectively compensating for key shortcomings of existing visual representations in VLAs. Building on these observations, we introduce JEPA-VLA, a simple yet effective approach that adaptively integrates predictive embeddings into existing VLAs. Our experiments demonstrate that JEPA-VLA yields substantial performance gains across a range of benchmarks, including LIBERO, LIBERO-plus, RoboTwin2.0, and real-robot tasks.

cs.CV↗

arXiv2026

MARTI-MARS$^2$: Scaling Multi-Agent Self-Search via Reinforcement Learning for Code Generation

While the complex reasoning capability of Large Language Models (LLMs) has attracted significant attention, single-agent systems often encounter inherent performance ceilings in complex tasks such as code generation. Multi-agent collaboration offers a promising avenue to transcend these boundaries. However, existing frameworks typically rely on prompt-based test-time interactions or multi-role configurations trained with homogeneous parameters, limiting error correction capabilities and strategic diversity. In this paper, we propose a Multi-Agent Reinforced Training and Inference Framework with Self-Search Scaling (MARTI-MARS2), which integrates policy learning with multi-agent tree search by formulating the multi-agent collaborative exploration process as a dynamic and learnable environment. By allowing agents to iteratively explore and refine within the environment, the framework facilitates evolution from parameter-sharing homogeneous multi-role training to heterogeneous multi-agent training, breaking through single-agent capability limits. We also introduce an efficient inference strategy MARTI-MARS2-T+ to fully exploit the scaling potential of multi-agent collaboration at test time. We conduct extensive experiments across varied model scales (8B, 14B, and 32B) on challenging code generation benchmarks. Utilizing two collaborating 32B models, MARTI-MARS2 achieves 77.7%, outperforming strong baselines like GPT-5.1. Furthermore, MARTI-MARS2 reveals a novel scaling law: shifting from single-agent to homogeneous multi-role and ultimately to heterogeneous multi-agent paradigms progressively yields higher RL performance ceilings, robust TTS capabilities, and greater policy diversity, suggesting that policy diversity is critical for scaling intelligence via multi-agent reinforcement learning.

cs.LG↗

arXiv2026

CompEvent: Complex-valued Event-RGB Fusion for Low-light Video Enhancement and Deblurring

Low-light video deblurring poses significant challenges in applications like nighttime surveillance and autonomous driving due to dim lighting and long exposures. While event cameras offer potential solutions with superior low-light sensitivity and high temporal resolution, existing fusion methods typically employ staged strategies, limiting their effectiveness against combined low-light and motion blur degradations. To overcome this, we propose CompEvent, a complex neural network framework enabling holistic full-process fusion of event data and RGB frames for enhanced joint restoration. CompEvent features two core components: 1) Complex Temporal Alignment GRU, which utilizes complex-valued convolutions and processes video and event streams iteratively via GRU to achieve temporal alignment and continuous fusion; and 2) Complex Space-Frequency Learning module, which performs unified complex-valued signal processing in both spatial and frequency domains, facilitating deep fusion through spatial structures and system-level characteristics. By leveraging the holistic representation capability of complex-valued neural networks, CompEvent achieves full-process spatiotemporal fusion, maximizes complementary learning between modalities, and significantly strengthens low-light video deblurring capability. Extensive experiments demonstrate that CompEvent outperforms SOTA methods in addressing this challenging task.

cs.CV↗

arXiv2026

Are Your Generated Instances Truly Useful? GenBench-MILP: A Benchmark Suite for MILP Instance Generation

The proliferation of machine learning-based methods for Mixed-Integer Linear Programming (MILP) instance generation has surged, driven by the need for diverse training datasets. However, a critical question remains: Are these generated instances truly useful and realistic? Current evaluation protocols often rely on superficial structural metrics or simple solvability checks, which frequently fail to capture the true computational complexity of real-world problems. To bridge this gap, we introduce GenBench-MILP, a comprehensive benchmark suite designed for the standardized and objective evaluation of MILP generators. Our framework assesses instance quality across four key dimensions: mathematical validity, structural similarity, computational hardness, and utility in downstream tasks. A distinctive innovation of GenBench-MILP is the analysis of solver-internal features -- including root node gaps, heuristic success rates, and cut plane usage. By treating the solver's dynamic behavior as an expert assessment, we reveal nuanced computational discrepancies that static graph features miss. Our experiments on instance generative models demonstrate that instances with high structural similarity scores can still exhibit drastically divergent solver interactions and difficulty levels. By providing this multifaceted evaluation toolkit, GenBench-MILP aims to facilitate rigorous comparisons and guide the development of high-fidelity instance generators.

cs.LG↗

arXiv2026

Lagged backward-compatible physics-informed neural networks for unsaturated soil consolidation analysis

This study develops a Lagged Backward-Compatible Physics-Informed Neural Network (LBC-PINN) for simulating and inverting one-dimensional unsaturated soil consolidation under long-term loading. To address the challenges of coupled air and water pressure dissipation across multi-scale time domains, the framework integrates logarithmic time segmentation, lagged compatibility loss enforcement, and segment-wise transfer learning. In forward analysis, the LBC-PINN with recommended segmentation schemes accurately predicts pore air and pore water pressure evolution. Model predictions are validated against finite element method (FEM) results, with mean absolute errors below 1e-2 for time durations up to 1e10 seconds. A simplified segmentation strategy based on the characteristic air-phase dissipation time improves computational efficiency while preserving predictive accuracy. Sensitivity analyses confirm the robustness of the framework across air-to-water permeability ratios ranging from 1e-3 to 1e3.

cs.LG↗

arXiv2026

ASTER: Agentic Scaling with Tool-integrated Extended Reasoning

Reinforcement learning (RL) has emerged as a dominant paradigm for eliciting long-horizon reasoning in Large Language Models (LLMs). However, scaling Tool-Integrated Reasoning (TIR) via RL remains challenging due to interaction collapse: a pathological state where models fail to sustain multi-turn tool usage, instead degenerating into heavy internal reasoning with only trivial, post-hoc code verification. We systematically study three questions: (i) how cold-start SFT induces an agentic, tool-using behavioral prior, (ii) how the interaction density of cold-start trajectories shapes exploration and downstream RL outcomes, and (iii) how the RL interaction budget affects learning dynamics and generalization under varying inference-time budgets. We then introduce ASTER (Agentic Scaling with Tool-integrated Extended Reasoning), a framework that circumvents this collapse through a targeted cold-start strategy prioritizing interaction-dense trajectories. We find that a small expert cold-start set of just 4K interaction-dense trajectories yields the strongest downstream performance, establishing a robust prior that enables superior exploration during extended RL training. Extensive evaluations demonstrate that ASTER-4B achieves state-of-the-art results on competitive mathematical benchmarks, reaching 90.0% on AIME 2025, surpassing leading frontier open-source models, including DeepSeek-V3.2-Exp.

cs.CL↗

arXiv2026

Learnable Permutation for Structured Sparsity on Transformer Models

Structured sparsity has emerged as a popular model pruning technique, widely adopted in various architectures, including CNNs, Transformer models, and especially large language models (LLMs) in recent years. A promising direction to further improve post-pruning performance is weight permutation, which reorders model weights into patterns more amenable to pruning. However, the exponential growth of the permutation search space with the scale of Transformer architectures forces most methods to rely on greedy or heuristic algorithms, limiting the effectiveness of reordering. In this work, we propose a novel end-to-end learnable permutation framework. Our method introduces a learnable permutation cost matrix to quantify the cost of swapping any two input channels of a given weight matrix, a differentiable bipartite matching solver to obtain the optimal binary permutation matrix given a cost matrix, and a sparsity optimization loss function to directly optimize the permutation operator. We extensively validate our approach on vision and language Transformers, demonstrating that our method achieves state-of-the-art permutation results for structured sparsity.

cs.LG↗

arXiv2026

Synthesis of Monolayer Ice on a Hydrophobic Metal Surface

Understanding water-metal interactions is central to disciplines spanning catalysis, electrochemistry, and atmospheric science. Monolayer ice phases are well established on hydrophilic surfaces, where strong water-substrate interactions stabilize ordered hydrogen-bond networks. In contrast, their formation on hydrophobic metals has been deemed ther-modynamically unfavourable, with water typically assembling into amorphous films, three-dimensional crystallites, or interlocked bilayer ice. Here, we demonstrate the synthesis of a monolayer ice phase on the hydrophobic Au(111) surface using a low-energy-electron-assisted growth method. Combined experimental characterizations including low-energy electron diffraction, angle-resolved photoemission spectroscopy, and X-ray photoelectron spectroscopy, complemented by first-principles calculations, prove that the monolayer ice phase composes of intact water molecules. This approach provides a generalizable strategy for stabilizing ordered two-dimensional ice on inert substrates and offers new insight into the interplay between water and low-energy electrons at hydrophobic interfaces.

cond-mat.mtrl-sci↗

arXiv2026

Performance Bounds of Joint Detection with Kalman Filtering and Channel Decoding for Wireless Networked Control Systems

The joint detection uses Kalman filtering (KF) to estimate the prior probability of control outputs to assist channel decoding. In this paper, we regard the joint detection as maximum a posteriori (MAP) decoding and derive the lower and upper bounds based on the pairwise error probability considering system interference, quantization interval, and weight distribution. We first derive the limiting bounds as the signal-to-noise ratio (SNR) goes to infinity and the system interference goes to zero. Then, we construct an infinite-state Markov chain to describe the consecutive packet losses of the control systems to derive the MAP bounds. Finally, the MAP bounds are approximated as the bounds of the transition probability from the state with no packet loss to the state with consecutive single packet loss. The simulation results show that the MAP performance of $\left(64,16\right)$ polar code and 16-bit CRC coincides with the limiting upper bound as the SNR increases and has $3.0$dB performance gain compared with the normal approximation of the finite block rate at block error rate $10^{-3}$.

cs.IT↗

arXiv2026

Histopathology-centered Computational Evolution of Spatial Omics: Integration, Mapping, and Foundation Models

Spatial omics (SO) technologies enable spatially resolved molecular profiling, while hematoxylin and eosin (H&E) imaging remains the gold standard for morphological assessment in clinical pathology. Recent computational advances increasingly place H&E images at the center of SO analysis, bridging morphology with transcriptomic, proteomic, and other spatial molecular modalities, and pushing resolution toward the single-cell level. In this survey, we systematically review the computational evolution of SO from a histopathology-centered perspective and organize existing methods into three paradigms: integration, which jointly models paired multimodal data; mapping, which infers molecular profiles from H&E images; and foundation models, which learn generalizable representations from large-scale spatial datasets. We analyze how the role of H&E images evolves across these paradigms from spatial context to predictive anchor and ultimately to representation backbone in response to practical constraints such as limited paired data and increasing resolution demands. We further summarize actionable modeling directions enabled by current architectures and delineate persistent gaps driven by data, biology, and technology that are unlikely to be resolved by model design alone. Together, this survey provides a histopathology-centered roadmap for developing and applying computational frameworks in SO.

q-bio.GN↗

arXiv2026

What Is The Best 3D Scene Representation for Robotics? From Geometric to Foundation Models

In this paper, we provide a comprehensive overview of existing scene representation methods for robotics, covering traditional representations such as point clouds, voxels, signed distance functions (SDF), and scene graphs, as well as more recent neural representations like Neural Radiance Fields (NeRF), 3D Gaussian Splatting (3DGS), and the emerging Foundation Models. While current SLAM and localization systems predominantly rely on sparse representations like point clouds and voxels, dense scene representations are expected to play a critical role in downstream tasks such as navigation and obstacle avoidance. Moreover, neural representations such as NeRF, 3DGS, and foundation models are well-suited for integrating high-level semantic features and language-based priors, enabling more comprehensive 3D scene understanding and embodied intelligence. In this paper, we categorized the core modules of robotics into five parts (Perception, Mapping, Localization, Navigation, Manipulation). We start by presenting the standard formulation of different scene representation methods and comparing the advantages and disadvantages of scene representation across different modules. This survey is centered around the question: What is the best 3D scene representation for robotics? We then discuss the future development trends of 3D scene representations, with a particular focus on how the 3D Foundation Model could replace current methods as the unified solution for future robotic applications. The remaining challenges in fully realizing this model are also explored. We aim to offer a valuable resource for both newcomers and experienced researchers to explore the future of 3D scene representations and their application in robotics. We have published an open-source project on GitHub and will continue to add new works and technologies to this project.

cs.RO↗

arXiv2026

SRFlow: A Dataset and Regularization Model for High-Resolution Facial Optical Flow via Splatting Rasterization

Facial optical flow supports a wide range of tasks in facial motion analysis. However, the lack of high-resolution facial optical flow datasets has hindered progress in this area. In this paper, we introduce Splatting Rasterization Flow (SRFlow), a high-resolution facial optical flow dataset, and Splatting Rasterization Guided FlowNet (SRFlowNet), a facial optical flow model with tailored regularization losses. These losses constrain flow predictions using masks and gradients computed via difference or Sobel operator. This effectively suppresses high-frequency noise and large-scale errors in texture-less or repetitive-pattern regions, enabling SRFlowNet to be the first model explicitly capable of capturing high-resolution skin motion guided by Gaussian splatting rasterization. Experiments show that training with the SRFlow dataset improves facial optical flow estimation across various optical flow models, reducing end-point error (EPE) by up to 42% (from 0.5081 to 0.2953). Furthermore, when coupled with the SRFlow dataset, SRFlowNet achieves up to a 48% improvement in F1-score (from 0.4733 to 0.6947) on a composite of three micro-expression datasets. These results demonstrate the value of advancing both facial optical flow estimation and micro-expression recognition.

cs.CV↗

arXiv2026

Adversarial Contrastive Learning for LLM Quantization Attacks

Model quantization is critical for deploying large language models (LLMs) on resource-constrained hardware, yet recent work has revealed severe security risks that benign LLMs in full precision may exhibit malicious behaviors after quantization. In this paper, we propose Adversarial Contrastive Learning (ACL), a novel gradient-based quantization attack that achieves superior attack effectiveness by explicitly maximizing the gap between benign and harmful responses probabilities. ACL formulates the attack objective as a triplet-based contrastive loss, and integrates it with a projected gradient descent two-stage distributed fine-tuning strategy to ensure stable and efficient optimization. Extensive experiments demonstrate ACL's remarkable effectiveness, achieving attack success rates of 86.00% for over-refusal, 97.69% for jailbreak, and 92.40% for advertisement injection, substantially outperforming state-of-the-art methods by up to 44.67%, 18.84%, and 50.80%, respectively.

cs.CR↗

arXiv2026

DiffBench Meets DiffAgent: End-to-End LLM-Driven Diffusion Acceleration Code Generation

Diffusion models have achieved remarkable success in image and video generation. However, their inherently multiple step inference process imposes substantial computational overhead, hindering real-world deployment. Accelerating diffusion models is therefore essential, yet determining how to combine multiple model acceleration techniques remains a significant challenge. To address this issue, we introduce a framework driven by large language models (LLMs) for automated acceleration code generation and evaluation. First, we present DiffBench, a comprehensive benchmark that implements a three stage automated evaluation pipeline across diverse diffusion architectures, optimization combinations and deployment scenarios. Second, we propose DiffAgent, an agent that generates optimal acceleration strategies and codes for arbitrary diffusion models. DiffAgent employs a closed-loop workflow in which a planning component and a debugging component iteratively refine the output of a code generation component, while a genetic algorithm extracts performance feedback from the execution environment to guide subsequent code refinements. We provide a detailed explanation of the DiffBench construction and the design principles underlying DiffAgent. Extensive experiments show that DiffBench offers a thorough evaluation of generated codes and that DiffAgent significantly outperforms existing LLMs in producing effective diffusion acceleration strategies.

cs.CV↗

arXiv2026

Ratio-Variance Regularized Policy Optimization for Efficient LLM Fine-tuning

On-policy reinforcement learning (RL), particularly Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), has become the dominant paradigm for fine-tuning large language models (LLMs). While policy ratio clipping stabilizes training, this heuristic hard constraint incurs a fundamental cost: it indiscriminately truncates gradients from high-return yet high-divergence actions, suppressing rare but highly informative "eureka moments" in complex reasoning. Moreover, once data becomes slightly stale, hard clipping renders it unusable, leading to severe sample inefficiency. In this work, we revisit the trust-region objective in policy optimization and show that explicitly constraining the \emph{variance (second central moment) of the policy ratio} provides a principled and smooth relaxation of hard clipping. This distributional constraint stabilizes policy updates while preserving gradient signals from valuable trajectories. Building on this insight, we propose $R^2VPO$ (Ratio-Variance Regularized Policy Optimization), a novel primal-dual framework that supports stable on-policy learning and enables principled off-policy data reuse by dynamically reweighting stale samples rather than discarding them. We extensively evaluate $R^2VPO$ on fine-tuning state-of-the-art LLMs, including DeepSeek-Distill-Qwen-1.5B and the openPangu-Embedded series (1B and 7B), across challenging mathematical reasoning benchmarks. Experimental results show that $R^2VPO$ consistently achieves superior asymptotic performance, with average relative gains of up to 17% over strong clipping-based baselines, while requiring approximately 50% fewer rollouts to reach convergence. These findings establish ratio-variance control as a promising direction for improving both stability and data efficiency in RL-based LLM alignment.

cs.LG↗

arXiv2026

Dual LoRA: Enhancing LoRA with Magnitude and Direction Updates

Low-rank adaptation (LoRA) is one of the most popular methods among parameter-efficient fine-tuning (PEFT) methods to adapt pre-trained large language models (LLMs) to specific downstream tasks. However, the model trained based on LoRA often has an unsatisfactory performance due to its low-rank assumption. In this paper, we propose a novel method called Dual LoRA to improve the performance by incorporating an inductive bias into the original LoRA. Specifically, we separate low-rank matrices into two groups: the magnitude group to control whether or not and how far we should update a parameter and the direction group to decide whether this parameter should move forward or backward, to better simulate the parameter updating process of the full fine-tuning based on gradient-based optimization algorithms. We show that this can be simply achieved by adding a ReLU function to the magnitude group and a sign function to the direction group. We conduct several experiments over a wide range of NLP tasks, including natural language understanding (NLU) and commonsense reasoning datasets on RoBERTa, DeBERTa, and LLaMA-1/2/3 as baseline models. The results show that we consistently outperform LoRA and its state-of-the-art variants with the same number of trainable parameters.

cs.CL↗

arXiv2025

Sequence to Sequence Reward Modeling: Improving RLHF by Language Feedback

Aligning the behavior of Large language models (LLMs) with human intentions and values remains a critical challenge. Reinforcement learning from human feedback (RLHF) aligns LLMs by training a reward model (RM) on human preferences and fine-tuning the LLMs to maximize RM feedback. Despite its effectiveness and popularity, RLHF is prone to biased local optimization. It means RM fails to provide feedback that accurately aligns with human preference, causing LLMs to explore unexpected generalizations, and failing to achieve alignment objectives. To mitigate this issue, we propose a novel \textit{sequence-to-sequence (seq2seq) reward modeling} method. Its key insight is that learning from language feedback rather than scalar feedback improves RLHF without additional annotations. We replaced the reward modeling target from binary maximum likelihood estimation (MLE) with sequence MLE. This method enables richer and fine-grained language feedback without additional annotations, models, or training stages. Our experiments demonstrated its effectiveness, specifically, reducing the refusal-to-response paradigm in single-turn safety dialogues and the long-response bias in text summarization tasks. We provide further analysis that seq2seq RM improves RLHF performance across 2B and 7B LLMs on 3 NLP tasks, achieving an average win rate of 76.9\%. We further show that seq2seq RM can still improve the performance of RLHF under out-of-distribution prompts.

cs.CL↗