arXiv subjects

Zihan Lin

Publications and source records attributed to Zihan Lin.

At least 19 recordsLinked to original sources

arXiv2026

Physically-Induced Atmospheric Adversarial Perturbations: Enhancing Transferability and Robustness in Remote Sensing Image Classification

Adversarial attacks pose a severe threat to the reliability of deep learning models in remote sensing (RS) image classification. Most existing methods rely on direct pixel-wise perturbations, failing to exploit the inherent atmospheric characteristics of RS imagery or survive real-world image degradations. In this paper, we propose FogFool, a physically plausible adversarial framework that generates fog-based perturbations by iteratively optimizing atmospheric patterns based on Perlin noise. By modeling fog formations with natural, irregular structures, FogFool generates adversarial examples that are not only visually consistent with authentic RS scenes but also deceptive. By leveraging the spatial coherence and mid-to-low-frequency nature of atmospheric phenomena, FogFool embeds adversarial information into structural features shared across diverse architectures. Extensive experiments on two benchmark RS datasets demonstrate that FogFool achieves superior performance: not only does it exceed in white-box settings, but also exhibits exceptional black-box transferability (reaching 83.74% TASR) and robustness against common preprocessing-based defenses such as JPEG compression and filtering. Detailed analyses, including confusion matrices and Class Activation Map (CAM) visualizations, reveal that our atmospheric-driven perturbations induce a universal shift in model attention. These results indicate that FogFool represents a practical, stealthy, and highly persistent threat to RS classification systems, providing a robust benchmark for evaluating model reliability in complex environments.

cs.CV↗

arXiv2026

S$^3$R: Fast and Faithful RL Rollouts via Self-Sparse Speculative Decoding

Fast rollout generation for long-context reinforcement learning with verifiable rewards (RLVR) requires reducing KV-cache traffic without changing the response groups used for policy optimization. In our profiled RLVR workloads, attention accounts for 72--76\% of autoregressive decoding kernel time. Sparse attention reduces these reads, but sampling directly from a sparse policy changes the response groups used for learning. For group-relative optimization, we show that even exact response-wise reweighting does not generally recover the expected dense-policy update because the advantage of each response depends on the rewards of other group members. We propose Self-Sparse Speculative Rollouts (S$^3$R). It uses sparse attention to propose tokens, then generates responses through full-attention verification and exact speculative sampling. Drafting and verification share the same policy parameter, eliminating the need to train or update a separate drafter. The resulting response groups follow the dense-policy distribution. Under S$^3$R, we characterize speedup conditions and the costs of drafting and verification. We found that dense verification still requires the complete historical KV cache. A KV-aware scheduler therefore limits KV-cache repeated offloading and restoration while maintaining rollout parallelism. % OR thrashing In a matched ablation with an aggressive sparse drafting budget, S$^3$R maintains stable training while direct sparse sampling with training-side correction becomes unstable. On long-context mathematical reasoning, S$^3$R achieves $1.39\times~1.35\times$ the end-to-end training throughput of autoregressive rollouts on Qwen3 model, with comparable task performance.

cs.LG↗

arXiv2026

Coupled-cluster molecular properties across the main group that extrapolate beyond training size

Coupled-cluster theory defines the accuracy standard for molecular electronic-structure properties but scales too steeply for routine application, whereas density-functional theory is affordable yet systematically biased. We resolve this trade-off with a single equivariant network, HARP (Hamiltonian Read-out for Properties), that predicts an effective one-electron Hamiltonian from one inexpensive B3LYP/def2-SVP calculation and derives a broad suite of properties from it (energy, optical gap, dipole, quadrupole, polarizability, Mulliken atomic charges, and Mayer bond orders) at coupled-cluster accuracy across nine main-group elements, including the under-served phosphorus, sulfur, and chlorine chemistries. The model is trained on a new in-house dataset of multi-property labels computed at the CCSD(T) level for all nine elements. On a held-out test set, it reduces the error of every property by a factor of 3.8 to 270 relative to semi-local, hybrid, and double-hybrid DFT (referenced to composite CCSD(T)/cc-pVTZ), while adding only ~0.1 s wall time per molecule, delivering coupled-cluster-quality predictions at the cost of a single DFT calculation. Critically, deriving every property from a predicted Hamiltonian rather than pooling per-atom features builds the correct size-scaling into the model architecture: on pi-conjugated oligothiophenes it matches finite-field CCSD polarizability to ~1% and the EOM-CCSD optical gap to ~3% at the largest sizes where those references remain affordable (44 and 37 atoms, where a single CCSD field point already costs ~500x the model's entire inference) and extrapolates the corrected trends to 58-atom chains, a regime where pooling-based architectures fail by construction. Accurate extrapolation is therefore set by the model's inductive bias rather than by the training data.

physics.chem-ph↗

arXiv2026

GaugeVLM: Structuring Spatial Supervision with Measured Geometric Interventions

Vision-language models (VLMs) can contradict themselves across views of the same spatial relation and fail to respond when that relation changes. Addressing these failures requires supervision that captures error magnitude and geometric dependencies across observations, both of which remain implicit in training on individual answers or ordinal preferences. Therefore, we introduce GaugeVLM, which makes this structure explicit through controlled object and camera interventions in explicit 3D scenes, producing linked observations with measured differences between spatial relations and shared truths across views. To translate this structure into learning signals, its core objective, GaugeDPO, converts measured errors into preference margins, directly supervises correct canonical rankings across views, and links intervention-induced answer-odds contrasts to measured relation changes with view-specific scales. Our analysis bounds canonical prediction error and establishes that the cross-view and intervention constraints can be jointly satisfied. Empirically, GaugeVLM improves all 10 established spatial metrics over supervised fine-tuning across three VLM backbones, with the main 7B model gaining 15.0 and 18.9 percentage points on MSMU distance and QSpatial+, respectively. These gains also extend to autonomous driving and embodied reasoning, demonstrating the robust generalization across domains.

cs.CV↗

arXiv2026

TGRL: Temperature-Grouped Reinforcement Learning for Efficient Exploration in LLMs

Efficient exploration often remains a central bottleneck in reinforcement learning with verifiable rewards (RLVR). Although temperature control and test-time scaling strategies can increase rollout diversity of large language models (LLMs), they either expand the sample budget at rollout time or leave the benefit of exploration unquantified. To this end, we propose Temperature-Grouped Reinforcement Learning (TGRL), which turns temperature-induced diversity into an explicit training signal. For each prompt, TGRL partitions its rollout group into low- and high-temperature subsets, estimates exploration gain through their reward contrast, and allocates this group-level signal as token-level credit using Jensen--Shannon (JS) divergence between the corresponding temperature-scaled next-token distributions induced by the same logits. Notably, TGRL reaches equivalent accuracy up to 36% faster than strong RLVR baselines without expanding the rollout budget. Across 11 benchmarks from diverse domains, TGRL broadly improves over strong RLVR baselines: it improves the six-benchmark math average by 1.6% at 32B, raises CodeForces rating by 196.7 points and LiveCodeBench Pass@16 by 4.4%, and improves ALFWorld/WebShop success rates by 6.3%/4.9%. Comprehensive ablations and wall-clock analysis confirm the efficacy of all proposed components. Code is available at https://github.com/1229095296/TGRL/tree/main.

cs.LG↗

arXiv2026

What Variation Identifies Payoffs in a Dynamic Game?

Observed choice in a dynamic game mixes current profit with continuation value. A rival adds a second problem: the same comparison averages over the rival's equilibrium policy. Changing the primitive transition rewrites continuation technology; changing the rival's Markov policy, holding that law fixed, rewrites the mixture over rival-contingent payoffs. The two are not substitutes. For a rival-feature payoff of rank $K$, rank identification up to location requires $\Ephi=\lceil(MK-1)/(M-1)\rceil$ policy environments, and a second kernel when payoffs are saturated. Rank can still be restored by arbitrarily small policy differences. Independent private shocks force mixed rival actions to factor, so a payoff that depends jointly on $d$ rivals is visible only at order $η^{d}$ near a common interior baseline. Either rank fails or the smallest identified singular value is at most $κη^{\dPhi}$, independently of how many kernels are stacked. Oracle-GLS variance in that direction vanishes only if $nη^{2\dPhi}$ diverges. An Anderson--Rubin set that carries first-stage error in the design matrix covers without a vanishing-risk condition. On U.S.\ airline entry, even among rank-identified directions, the most favorable rival-dependent contrast is several times wider than observed behavior.

econ.EM↗

arXiv2026

Policy Gains or Household Need? The Allocation Logic of China's Dibao Program

When social assistance is scarce, should it prioritize households in greatest need or those expected to benefit most? Using panel data from the China Household Finance Survey, we distinguish allocation principles by combining predicted policy gains with entry into China's Minimum Living Standard Guarantee (Dibao). We estimate heterogeneous predicted gains in consumption and education and then recover the conditional priorities revealed by recipient selection. Predicted gains explain little of allocation: a Shapley decomposition attributes 96.9\% of the improvement in allocation fit to household priorities and 3.1\% to predicted gains. Lower income, lower education of the household head, and elderly presence consistently predict higher priority across supported outcome-value specifications. Holding local program capacity fixed and removing household-priority differences while retaining the full model's estimated outcome values, the resulting ranking overlaps with recipients by only 10.9\%, implying 89.1\% recipient churn. These findings show that Dibao allocation is more closely aligned with household circumstances associated with poverty and vulnerability than with policy gains predictable from the outcomes and information observed in our data, underscoring the distinction between distributional and impact targeting in social assistance.

econ.EM↗

arXiv2026

One Step, One Lead: Mitigating Higher-Order Interference in Multi-Domain Reinforcement Learning via Cross-Step Control

Reinforcement learning (RL) across multiple domains can broaden the reasoning capabilities of large language models (LLMs), yet joint training often degrades individual-domain performance and can destabilize optimization. Existing work typically diagnoses such interference from a single-step view using first-order gradient alignment or curvature-based proxies. We show that this view can miss a critical form of sequential interference: same-point domain gradients may remain nearly orthogonal even when consecutive realized updates partially reverse one another in output space. We further show that consecutive token log-probability footprints recover this interaction directly from adjacent checkpoints as a local second-order interaction in output space, without explicitly reconstructing same-step curvature. Building on this insight, we propose OSOL, which designates a focus domain at each iteration, uses the preceding checkpoint footprint to rank token-level rebound risk, and applies a drift-ranked, adaptively scaled correction within the standard GRPO update. Our analysis shows that this correction suppresses the targeted cross-step output backtracking component. Controlled studies further show that cross-step backtracking is more strongly associated with subsequent task damage than same-point gradient diagnostics, while the preceding footprint ranks future rebound risk more accurately than Hessian-based proxies. On Qwen3-30B-A3B, OSOL reaches a domain-macro average of 0.4822, improving by 5.7% over the strongest compared baseline, without explicit higher-order differentiation.

cs.LG↗

arXiv2026

When Not to Imitate: Boundary-Aware Skill Memory for Reliable Tool-Use LLM Agents

Extracting skills from past successes is critical for the efficient evolution of Large Language Model (LLM) agents. Prevailing agent self-evolution paradigms typically rely on a core assumption: equipping LLMs with skill memories derived from successful trajectories will monotonically improve their problem-solving capabilities. However, probe analyses reveal that extracting skills solely from successful trajectories traps the model in a \textbf{Skill Imitation Trap}. For tasks that resemble past successes but require different tools, retrieving more skills paradoxically increases the model's confidence in wrong tool calls---procedure skills raise the wrong-tool margin by $47\%$ over a memory-free baseline. To overcome this limitation, we propose \textbf{Boundary-Aware Skill Memory} (BASM), which augments each skill with explicit boundary fields---applicability conditions, risk cues, avoidance rules, and recovery notes. These fields transform each retrieved skill from an unconditional action template into state-conditioned guidance: the agent applies the skill when its conditions hold, suppresses inapplicable tool calls when they do not, and issues targeted repairs when execution fails. Across three agent benchmarks and four model scales, BASM consistently outperforms success-distilled skill-memory baselines: it improves task success rate by up to $23.8\%$ on AppWorld, accuracy by up to $5.0\%$ on BFCL, and reduces attack success rate by $4.6\%$ on AgentDojo, while simultaneously reducing average AppWorld steps by up to $6.6\%$ relative to the memory-free baseline.

cs.CL↗

arXiv2026

LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation

Large Language Models (LLMs) have shown great promise in recommendation due to superior reasoning abilities. However, existing methods mainly rely on explicit Chain-of-Thought (CoT), resulting in verbose reasoning texts and inefficient response times. latent reasoning aims to balance efficiency by thinking within a continuous latent space, yet it faces two major challenges: (1) Lack of Fine-grained Supervision: Latent reasoning relies solely on feedback from the final labels, providing sparse supervisory signals that struggle to effectively guide the optimization of multiple hidden reasoning steps. (2) Single Reasoning Path: The deterministic nature of latent reasoning impedes the exploration of users' diverse interests and preferences, thereby limiting the recommendation capabilities of LLMs. To address these issues, we propose \textbf{$LaRec$}, an efficient generative recommendation framework designed to unleash the potential of latent reasoning in LLMs. $LaRec$ consists of two core stages: First, we design Latent Pre-training that empowers LLMs with latent reasoning capabilities by providing rich supervisory signals to the latent space reasoning via step-level alignment and process direction alignment. Second, we introduce Personalized RL-tuning. Specifically, we construct a personalized Gaussian Mixture Distribution for each user based on their historical interests. By randomly sampling distinct reasoning starting points from this distribution during training, we guide the LLMs to traverse diverse reasoning paths within the latent space, enabling efficient exploration of user's multi-faceted interests. Experiments on multiple datasets show that $LaRec$ significantly outperforms existing baselines with comparable efficiency.

cs.IR↗

arXiv2026

Action-Aware Generative Sequence Modeling for Short Video Recommendation

With the rapid development of the Internet, users have increasingly higher expectations for the recommendation accuracy of online content consumption platforms. However, short videos often contain diverse segments, and users may not hold the same attitude toward all of them. Traditional binary-classification recommendation models, which treat a video as a single holistic entity, face limitations in accurately capturing such nuanced preferences. Considering that user consumption is a temporal process, this paper demonstrates that the timing of user actions can represent diverse intentions through statistical analysis and examination of action patterns. Based on this insight, we propose a novel modeling paradigm: Action-Aware Generative Sequence Network (A2Gen), which refines user actions along the temporal dimension and chains them into sequences for unified processing and prediction. First, we introduce the Context-aware Attention Module (CAM) to model action sequences enriched with item-specific contextual features. Building upon this, we develop the Hierarchical Sequence Encoder (HSE) to learn temporal action patterns from users' historical actions. Finally, through leveraging CAM, we design a module for action sequence generation: the Action-seq Autoregressive Generator (AAG). Extensive offline experiments on the Kuaishou's dataset and the Tmall public dataset demonstrate the superiority of our proposed model. Furthermore, through large-scale online A/B testing deployed on Kuaishou's platform, our model achieves significant improvements over baseline methods in multi-task prediction by leveraging sequential information. Specifically, it yields increases of 0.34% in user watch time, 8.1% in interaction rate, and 0.162% in overall user retention (LifeTime-7), leading to successful deployment across all traffic, serving over 400 million users every day.

cs.AI↗

arXiv2026

Agentic AI-RAN Empowering Synergetic Sensing, Communication, Computing, and Control

Future sixth-generation (6G) networks are expected to support low-altitude wireless networks (LAWNs), where unmanned aerial vehicles (UAVs) and aerial robots operate in highly dynamic three-dimensional environments under stringent latency, reliability, and autonomy requirements. In such scenarios, autonomous task execution at the network edge demands holistic coordination among sensing, communication, computing, and control (SC3) processes. Agentic Artificially Intelligent Radio Access Networks (Agentic AI-RAN) offer a promising paradigm by enabling the edge network to function as an autonomous decision-making entity for low-altitude agents with limited onboard resources. In this article, we propose a task-oriented Agentic AI-RAN architecture that enables SC3 task execution within a single edge node. The proposed architecture addresses the challenge of coordinating heterogeneous workloads in resource-constrained edge environments. To validate this framework, we prototype a representative low-altitude UAV system on a general-purpose Graphics Processing Unit (GPU) platform and evaluate it through an autonomous drone-navigation case study. The current prototype instantiates the platform-agnostic design through Multi-Instance GPU (MIG) partitioning and containerized deployment, providing physical resource isolation and coordinated execution between real-time communication and multimodal inference. Experimental results demonstrate low closed-loop latency, robust bidirectional communication, and stable performance under dynamic runtime conditions, highlighting the feasibility of the proposed framework for mission-critical low-altitude wireless networks in 6G.

eess.SY↗

arXiv2026

Spectroscopic fingerprints of a ferroaxial charge density wave

Unconventional charge density waves (CDWs) with complex order parameters can host exotic collective modes and non-trivial topologies. They have emerged as a new frontier in the study of quantum matter. Recent experiments on rare-earth tritellurides have reported evidence for a ferroaxial CDW through the detection of characteristic Raman modes. This phase, often regarded as a hidden order, has been recognized to arise from the coupling between charge and orbital degrees of freedom in these materials. Yet, spectroscopic insight into its underlying electronic structure and the explicit form of its order parameter symmetry has remained elusive. Here, we present results from linearly polarized angle-resolved photoemission spectroscopy (ARPES) and scanning tunneling microscopy (STM) measurements of the CDW phase in LaTe$_3$. Our ARPES measurements reveal a complex landscape of spectral gaps across the reconstructed Fermi surface, while our STM-based quasiparticle interference (QPI) mapping, enhanced through the selective deposition of atomic scattering centers, directly reveals an inter-orbital CDW with mixed $p_x$-$p_z$ orbital character. The detailed analysis of the QPI characteristics in terms of the order parameter symmetry within the orbital subspace of the Fermi surface suggests a mixed CDW phase with substantial ferroaxial component, which breaks all vertical mirror symmetries. More broadly, our work establishes a powerful spectroscopic pathway, based on scattering off individual atoms, for identifying and characterizing hidden, multi-component electronic orders in quantum materials using STM and ARPES measurements.

cond-mat.str-el↗

arXiv2026

CodecCap: High-Fidelity Codec-Inspired Residual Modeling for Dense Video Captioning

Existing video captioning methods struggle to balance visual fidelity and redundancy: holistic captions are compact but lose fine-grained evidence, whereas segment-wise captions improve coverage but introduce heavy redundancy. We propose CodecCap, a codec-inspired framework for high-fidelity dense video captioning. Analogous to video codecs, CodecCap represents videos using keyframe and residual captions. Keyframe captions exhaustively encode stable visual context, while residual captions capture temporally only localized actions, motions and changes. This effectively preserves fine-grained visual evidence while reducing redundant descriptions. To quantify the fidelity of captions, we introduce VidCapQA, a caption-then-QA benchmark with 1,000 questions across 14 capability dimensions. Results on VidCapQA show that captions directly generated by strong VLMs still miss many visual details, highlighting caption representation as a critical bottleneck. Experiments show that CodecCap significantly surpasses direct captioning with the same underlying VLMs, suggesting keyframe-residual captioning a way for high-fidelity video-language supervision. We further use CodecCap to construct CodecVDC-100K, a large-scale dense captioning dataset with anchor, residual, scene-level, and video-level supervision.

cs.CV↗

arXiv2026

VDEGaussian: Video Diffusion Enhanced 4D Gaussian Splatting for Dynamic Urban Scenes Modeling

Dynamic urban scene modeling is a rapidly evolving area with broad applications. While current approaches leveraging neural radiance fields or Gaussian Splatting have achieved fine-grained reconstruction and high-fidelity novel view synthesis, they still face significant limitations. These often stem from a dependence on pre-calibrated object tracks or difficulties in accurately modeling fast-moving objects from undersampled capture, particularly due to challenges in handling temporal discontinuities. To overcome these issues, we propose a novel video diffusion-enhanced 4D Gaussian Splatting framework. Our key insight is to distill robust, temporally consistent priors from a test-time adapted video diffusion model. To ensure precise pose alignment and effective integration of this denoised content, we introduce two core innovations: a joint timestamp optimization strategy that refines interpolated frame poses, and an uncertainty distillation method that adaptively extracts target content while preserving well-reconstructed regions. Extensive experiments demonstrate that our method significantly enhances dynamic modeling, especially for fast-moving objects, achieving an approximate PSNR gain of 2 dB for novel view synthesis over baseline approaches.

cs.CV↗

arXiv2026

ResRL: Boosting LLM Reasoning via Negative Sample Projection Residual Reinforcement Learning

Reinforcement Learning with Verifiable Rewards (RLVR) enhances reasoning of Large Language Models (LLMs) but usually exhibits limited generation diversity due to the over-incentivization of positive rewards. Although methods like Negative Sample Reinforcement (NSR) mitigate this issue by upweighting penalty from negative samples, they may suppress the semantic distributions shared between positive and negative responses. To boost reasoning ability without losing diversity, this paper proposes negative sample projection Residual Reinforcement Learning (ResRL) that decouples similar semantic distributions among positive and negative responses. We theoretically link Lazy Likelihood Displacement (LLD) to negative-positive head-gradient interference and derive a single-forward proxy that upper-bounds representation alignment to guide conservative advantage reweighting. ResRL then projects negative-token hidden representations onto an SVD-based low-rank positive subspace and uses projection residuals to modulate negative gradients, improving reasoning while preserving diversity and outperforming strong baselines on average across twelve benchmarks spanning Mathematics, Code, Agent Tasks, and Function Calling. Notably, ResRL surpasses NSR on mathematical reasoning by 9.4\% in Avg@16 and 7.0\% in Pass@128. Code is available at https://github.com/1229095296/ResRL.git.

cs.LG↗

arXiv2026

TimeMM: Time-as-Operator Spectral Filtering for Dynamic Multimodal Recommendation

Multimodal recommendation improves user modeling by integrating collaborative signals with heterogeneous item content. In real applications, user interests evolve over time and exhibit nonstationary dynamics, where different preference factors change at different rates. This challenge is amplified in multimodal settings because visual and textual cues can dominate decisions under different temporal regimes. Despite strong progress, most multimodal recommenders still rely on static interaction graphs or coarse temporal heuristics, which limits their ability to model continuous preference evolution with fine-grained temporal adaptation. To address these limitations, we propose TimeMM, a time-conditioned spectral filtering framework for dynamic multimodal recommendation. TimeMM instantiates Time-as-Operator by mapping interaction recency to a family of parametric temporal kernels that reweight edges on the user--item graph, producing component-specific representations without explicit eigendecomposition. To capture non-stationary interests, we introduce Adaptive Spectral Filtering that mixes the operator bank according to temporal context, yielding prediction-specific effective spectral responses. To account for modality-specific temporal sensitivity, we further propose Spectral-Aware Modality Routing that calibrates visual and textual contributions conditioned on the same temporal context. Finally, a ranking-space Spectral Diversity Regularization encourages complementary expert behaviors and prevents filter-bank collapse. Extensive experiments on real-world benchmarks demonstrate that TimeMM consistently outperforms state-of-the-art multimodal recommenders while maintaining linear-time scalability.

cs.IR↗

arXiv2026

Discovery of an odd-parity f-wave charge order in a kagome metal

The spontaneous breaking of symmetries is a cornerstone of physics, defining the phases of matter from the cosmological scale to the quantum realm. In condensed matter, electronic orders are classified by their behavior under fundamental symmetries like spatial inversion (parity). While even-parity orders, such as conventional superconductivity and charge density waves, are ubiquitous, their odd-parity counterparts--predicted to host exotic phenomena such as gapless quasiparticle excitations and novel collective modes--are comparatively elusive states of quantum matter. Here, using high-resolution scanning tunneling microscopy and angle-resolved photoemission spectroscopy on the kagome metal CsV$_3$Sb$_5$, we report the discovery of an inversion symmetry-breaking $f$-wave charge bond order. We show that this phase, which preserves translation symmetry, is stabilized by the spontaneous opening of a spectral gap at a previously overlooked Dirac point, providing a textbook condensed-matter realization of the Gross-Neveu model for dynamical mass generation and parity breaking. Intriguingly, this $f$-wave order is itself a intervening phase, vanishing abruptly below a temperature of 10\,K and pointing to a subsequent transition into a `hidden' electronic state that is invisible to local STM probes. Our findings establish odd-parity charge order as a novel phase of matter, here, embedded within the intricate hierarchy of correlated electronic orders on the kagome lattice.

cond-mat.str-el↗