arXiv subjects

Hongyu Liu

Publications and source records attributed to Hongyu Liu.

At least 55 records · Page 3Linked to original sources

arXiv2026

Emergent interweaved CDW unoccupied states in hole-doping LaTe2 with element substitution

Multiple CDW-ordered layered rare-earth tellurides have increasingly emerged as a research hotspot, owing to their unconventional CDW formation, high transition temperature, and confirmed existence of axial Higgs modes. Recently, interweaved CDW in LaTe2 and its element-substituted phase LaTe2-xSbx have been investigated through TEM and ARPES measurements, revealing their distinct origins. Nevertheless, several complex diffraction features observed in TEM patterns remain unelucidated. In this work, we carried out scanning tunneling microscopy (STM) on LaTe1.6Sb0.4 crystals at 9 K. Three interweaved CDW wave vectors, q1=8/11a*, q2=5/11a* and q3=3/11a* were observed, which are induced by hole doping in unoccupied states. The q1 and q3 are theoretically verified to be nesting vectors connecting px- and py- bands. Furthermore, the satellite spots relative to the main Bragg spots p corresponding to a 11-a-superlattice have also been detected. Our findings provide critical insights for further exploring the origin of the interweaved CDW in hole/electron doping materials.

cond-mat.str-el↗

arXiv2026

Localized gradient enhancement near anisotropic electromagnetic scatterers

This work investigates time-harmonic electromagnetic scattering governed by the Maxwell system, where bounded anisotropic scatterers are embedded in a homogeneous electromagnetic background. We focus on the localized enhancement of the gradients of the total electric and magnetic fields in small boundary-attached neighborhoods of finitely many prescribed points near boundaries of anisotropic electromagnetic scatterers. We show that, through a suitable construction of incident electromagnetic waves, the gradients of both the total electric field and the total magnetic field can be made arbitrarily large in these neighborhoods. The main strategy is based on the introduction of auxiliary boundary-attached electromagnetic neighborhoods and the associated electric and magnetic fields, which exhibit strong gradient variation near the prescribed points. Using the approximation property of Maxwell Herglotz wave functions, these auxiliary fields are then approximated by physically admissible incident waves in the neighborhood of the scatterers. Together with the well-posedness and continuous dependence of the anisotropic scattering problem, this implies that the corresponding scattered field can be controlled to be sufficiently weak in the relevant region. Consequently, the total field is dominated by the incident field near the prescribed points and inherits its large-gradient behavior. The result provides a theoretical mechanism for localized gradient enhancement in anisotropic electromagnetic scattering and may have implications for field concentration, high-resolution probing, and sensitivity analysis of electromagnetic responses in complex media.

math.AP↗

arXiv2026

OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators

We propose OPSD-V, an on-policy self-distillation paradigm for post-training few-step autoregressive (AR) video diffusion models. Existing few-step AR video generators can produce long videos with low latency, but still suffer from error accumulation and weakened motion dynamics during long autoregressive rollout. OPSD-V reduces long-horizon degradation while preserving the original few-step inference path. The key idea is to introduce real long-video data as temporal context during training and use it to provide dense trajectory-level supervision. Specifically, the student follows the exact inference-time rollout, generating each chunk conditioned on its own previously generated KV cache. In parallel, the teacher is evaluated at the same student-visited denoising states, but uses a cleaner AR-consistent temporal cache in which older history can be replaced by real-video context. This provides dense denoising-level corrective targets under on-policy AR cache dynamics, without changing the sampler, number of denoising steps, or inference-time cache mechanism. We apply OPSD-V to representative few-step AR video models, including Self-Forcing and LongLive. Experiments show consistent improvements in visual quality, motion dynamics, and VBenchLong scores. A user study with 10 participants comparing 20 video pairs shows that OPSD-V is preferred over the base models in 66.0% of overall-preference judgments (82.5% excluding ties).

cs.CV↗

arXiv2026

A Reliable Context-Aware and Temporal Planning Framework for Autonomous Driving

Safe operation of autonomous vehicles in dense urban traffic depends on perception and planning that remain reliable when onboard sensing is degraded. In real driving conditions, camera observations are frequently corrupted by occlusion, motion blur, illumination change, and sensor noise, and when such degraded observations are aggregated indiscriminately over time, trajectory planning becomes unstable and collision risk rises for both the ego vehicle and surrounding road users. Recent Bird's-Eye-View (BEV) approaches unify perception and planning through a shared spatial representation, but most fuse temporal information across frames without assessing the reliability of the underlying observations. We present a Reliable Context-Aware and Temporal Planning framework for Autonomous Driving (RCT-AD) that explicitly models feature quality and temporal consistency to support safer, more consistent planning. A Reliable Context Awareness module scores per-frame reliability and selectively retains trustworthy features through a quality-gated First-In-Last-Out (FILO) memory mechanism, reconstructing degraded observations from reliable historical context so that corrupted inputs do not destabilize the scene representation. A Temporal Trajectory Planner captures long-term dependencies and multi-agent interactions to produce smoother, safety-aware trajectories, while a joint detection-and-segmentation head injects semantic and motion cues into the shared BEV space to strengthen scene understanding. Experiments on the nuScenes autonomous driving benchmark show that RCT-AD improves perception accuracy, motion prediction, and planning robustness over recent end-to-end baselines, achieving 61.5 nuScenes Detection Score, 52.9 mean Average Precision, and 52.3 mean Intersection over Union, while maintaining competitive computational efficiency suitable for real-time deployment.

cs.RO↗

arXiv2026

Rise From The Ashes: LLM-based Static Analysis for Deep Learning Framework Bugs

Deep learning (DL) frameworks are critical AI infrastructures that often hide bugs with serious security implications. While dynamic approaches such as fuzzing are effective in uncovering these bugs, they require real test execution and incur high computational costs. Static analysis is a natural complement because it can detect bugs without runtime execution, offering fast and scalable testing. Unfortunately, there is still limited work targeting static analysis for DL frameworks due to their multilingual architectures and tensor-related program state. We present Phoenix, the first LLM-based static analysis technique for DL frameworks. Our key insight is that cross-language tensor flows in DL frameworks can be modeled, together with concrete code context, as a structured semantic bridge intermediate representation (SBIR) that LLMs can analyze for potential bugs in tensor semantic propagation. We implement this insight through a multi-agent workflow. A summarization agent first distills bug summaries from historical bug-fix patches and CWE rules. Guided by each summary, an extraction agent identifies bug-relevant repository symbols for code retrieval, and a generation agent synthesizes grounded SBIRs from the retrieved context. Finally, an analysis agent is leveraged to check SBIRs and report potential bugs. Our evaluation shows that Phoenix is a practical complement to dynamic DL framework testing for bug finding. To date, Phoenix has found 31 real new bugs in PyTorch for different heterogeneous hardware backends (Intel CPU, NVIDIA CUDA, and Apple MPS). Among them, 20 submitted bug-fixing patches have been merged into upstream.

cs.SE↗

arXiv2026

High performance construction materials fracture and high cycle fatigue assessment based on accelerated PF-CZM

Given the widespread application of high-performance materials in cyclic loaded infrastructure, a thorough understanding of the fracture and high-cycle fatigue behavior of high-performance materials is of critical importance. However, these behaviors predominantly rely on experimental methods, which are often costly, time-consuming and limited in generalizability. To address these issues, this paper proposes the constitution and high-cycle fatigue of Ultra-High Performance Concrete (UHPC) and High Strength Steel (HSS) within the Phase-Field Cohesive Zone Model (PF-CZM) framework. A generic stress-based failure criterion for UHPC is derived from the biaxial tensile test results to calculate crack driving force. Furthermore, a fatigue degradation function and an acceleration algorithm are integrated into the PF-CZM framework to enable efficient high-cycle fatigue simulations. In the acceleration algorithm, the envelope load is used to approximate the real cyclic load to avoid the simulation of each cycle, and the three-stage fatigue process is simulated through adjusting the cyclic increment adaptively. The proposed model successfully captures mixed-mode fracture and high-cycle fatigue behavior of UHPC and HSS, with validation provided through relevant experimental comparisons.

physics.app-ph↗

arXiv2026

A Measure-Consistent Operator Learning Method for Infinite-Dimensional Master Equations

Master equations in mean field game theory characterize feedback value functions that depend on time, state (space), and the population distribution. Their numerical approximation is challenging because the unknown is defined on a space of probability measures and the equation involves intrinsic measure derivatives and nonlocal population terms. This paper proposes a measure-consistent operator learning method (MCOL) for infinite-dimensional master equations. The population distribution is represented by an empirical measure and encoded through a symmetric pooling structure, so that the network input is built directly from the particles representing the measure. The same particles are used in the empirical quadrature of the nonlocal residual terms, avoiding additional quadrature grids or auxiliary integration points. A key feature is that the intrinsic derivative appearing in the residual is induced by the same measure-dependent representation that defines the approximation of the value function. Consequently, the value function, its measure derivative, and the empirical residual are tied to a common measure representation, leading to a structurally coupled value-derivative approximation. We also introduce an error decomposition separating neural approximation error from empirical discretization error. Numerical experiments on several master equations show that MCOL accurately approximates the value function, intrinsic measure derivatives, and feedback quantities, and remains robust under changes in the input measures.

math.NA↗

arXiv2026

Spectral structures of elastic-electromagnetic transmission eigenvalue problems

The time-harmonic elastic-electromagnetic interior transmission eigenvalue problem (EEITEP) arises when an elastic body becomes invisible to an incident electromagnetic wave. This spectral problem is typically non-elliptic and non-self-adjoint, making its analysis delicate. In this paper, we study the discreteness of transmission eigenvalues and the boundary localization of the associated eigenfunctions. For a general bounded Lipschitz domain, we prove that the set of positive transmission eigenvalues, if non-empty, is discrete with $\infty$ as its only possible accumulation point. For a radially symmetric domain, we demonstrate the existence of a sequence of transmission eigenvalues and derive their asymptotic behavior. We rigorously show that the corresponding transmission eigenfunctions exhibit boundary localization in their electromagnetic components, whereas the elastic displacement field remains globally distributed throughout the domain. Finally, we derive lower bounds for the $L^{\infty}$-norms of the electromagnetic gradients normalized by their $L^2$-norms, quantifying their blow-up behavior near the boundary along this sequence. These findings reveal a potential spectral mechanism for developing super-resolution imaging methods in elastic-electromagnetic scattering.

math.AP↗

arXiv2026

DyaPlex: Full-Duplex Speech-Motion Model for Dyadic Interaction

We present DyaPlex, a streaming, full-duplex speech-and-motion model designed for dyadic interaction. To capture the continuous and reciprocal nature of human communication, this full-duplex capability empowers the agent to simultaneously perceive and generate both speech and physical motion in a streaming fashion. At its core, our method leverages the strong priors of a foundational full-duplex speech model and integrates a novel motion pathway, thereby achieving fully synchronized multi-modal interaction. Specifically, we design a dual-tower Transformer architecture that preserves the zero-shot conversational reasoning of a frozen base speech model while constructing a deeply coupled, streaming motion pathway. By introducing a unified dyadic token interleaving mechanism and guiding cross-attention via a time-aligned speech-motion RoPE, our model effectively aligns autoregressive motions with rich latent speech features. Trained on the 4,000-hour Seamless Interaction dataset, our model effectively captures cross-speaker dependencies and establishes new state-of-the-art performance across both monadic and dyadic human interaction benchmarks.

cs.CV↗

arXiv2026

Accelerating Sparse Transformer Inference on GPU

Large language models (LLMs) are popular around the world due to their powerful understanding capabilities. As the core component of LLMs, accelerating Transformer through parallelization has gradually become a hot research topic. Mask layers introduce sparsity into Transformer to reduce calculations. However, previous works rarely focus on the performance optimization of sparse Transformer. In addition, current static operator fusion schemes fail to adapt to diverse application scenarios. To address the above problems, we propose STOF, a framework that incorporates optimizations for Sparse Transformer that enables flexible masking and Operator Fusion on GPU. For multi-head attention (MHA) structure, STOF maps the computation to row-wise or blockwise kernels with unique storage formats according to analytical modeling. For downstream operators, STOF maps the fusion scheme to compilation templates and determines the optimal running configuration through two-stage searching. The experimental results show that compared to the stateof-the-art work, STOF achieves maximum speedups of 1.6x in MHA computation and 1.4x in end-to-end inference.

cs.LG↗

arXiv2026

LongCat-Video-Avatar 1.5 Technical Report

Despite advances in audio-driven video generation, achieving commercial-grade stability remains challenging. We present LongCat-Video-Avatar 1.5, an upgraded open-source framework prioritizing systematic engineering and production-readiness over architectural novelty. By upgrading the audio encoder to Whisper Large and meticulously scaling our training recipes, v1.5 achieves accurate lip-synchronization, full-body temporal stability, and robust long-video generation with strict identity consistency. Through rigorous data curation and RLHF Training, the model readily generalizes to stylized domains such as anime and animals, and natively handles complex real-world conditions, such as multi-person interactions and object handling. Furthermore, addressing the practical demands of industrial deployment, we employ advanced step distillation to accelerate inference to an optimal 8 NFE, achieving a favorable trade-off between serving efficiency and visual fidelity. The superiority of our approach is validated through extensive quantitative metrics and a rigorous human evaluation conducted on a comprehensive benchmark of over 500 diverse test cases. Results show that v1.5 achieves competitive or superior performance compared to leading closed-source systems (e.g., HeyGen, OmniHuman 1.5, Kling Avatar 2.0) across human-likeness ratings and expert-level quality assessments on our benchmark. With its open-source release, LongCat-Video-Avatar 1.5 narrows the gap between academic research prototypes and commercial-grade deployment.

cs.CV↗

arXiv2026

Traceable Knowledge Graph Reasoning Enables LLM-Assisted Decision Support for Industrial VOCs in the Steel Industry

Key knowledge for steel-industry volatile organic compounds (VOCs) governance is scattered across unstructured scientific literature, making it difficult to integrate process, pollutant, and control-technology evidence and increasing the risk of hallucination when general large language models (LLMs) answer low-frequency industrial questions. Here we developed Chat-ISV, a knowledge graph (KG) enhanced multi-agent Q&A system that parses a curated steel-industry VOCs literature corpus, constructs a Neo4j KG with 27180 nodes and 81779 semantic edges, and combines prompt-constrained extraction, chunk-centered topology optimization, multi-agent routing, source-backtracking retrieval, local literature retrieval, open-domain knowledge access, and interactive subgraph visualization. Benchmark tests and 400 expert blind evaluations showed that topology optimization reduced isolated nodes from 57% to 4.08% and that Chat-ISV achieved high factual reliability, with 96.93% precision, 72.63% recall, an F1-score of 0.830, and a mean score of 1.69/2.00. By converting fragmented environmental-engineering literature into traceable, queryable, and decision-support-oriented knowledge, Chat-ISV establishes a scalable environmental-informatics paradigm for reliable LLM deployment and intelligent pollution-control decision support in specialized industrial domains.

cs.AI↗

arXiv2026

Determining evolutionary equations from a single passive boundary observation

We study inverse boundary problems for evolutionary PDEs using only a single passive boundary observation, where data from an unknown internal source propagate through an unknown medium without active inputs. The goal is the simultaneous recovery of coupled unknowns (sources and coefficients) from severely limited data. Unlike active methods with rich, structured inputs, passive observation poses two core challenges: minimal information and intrinsic coupling of multiple unknowns. Consequently, such problems remain largely open and unsystematically studied. We develop a unified framework based on integral identities, harmonic and microlocal analysis, and low-/high-frequency asymptotics. This approach yields the first systematic resolution for second-order hyperbolic, parabolic, and Schrödinger equations under a single coherent method. The key condition requires the measurement dataset's cardinality to exceed the unknowns' by at least one dimension, providing room to decouple unknowns and linearize the nonlinear inverse problem. Our unique identifiability results subsume all existing literature and cover more general configurations of practical interest. This framework complements classical theories and opens a promising new direction for future development.

math.AP↗

arXiv2026

Travel-time tomography from mean field game dynamics

Travel-time tomography seeks to recover a hidden environment from external measurements generated by propagation through an anomalous region. Standard formulations treat propagation as passive, so the environment influences observations mainly by bending paths or changing travel times. Many collective systems do not operate in that regime: observed arrivals are shaped by strategic motion, congestion, and environmental costs. We formulate this active setting through mean field games, in which the unknown environment enters the running cost through a spatial cost field and observations are read from the resulting population dynamics. This yields three contributions. First, it places propagation, observation generation, and inversion within one PDE-constrained model. Second, it clarifies why the inverse problem differs structurally from passive tomography: kinetic, congestion, and environmental effects are coupled endogenously and appear through space- and time-dependent local weights rather than externally chosen global coefficients. Third, it introduces a two-stage inversion pipeline that combines diffusion-based initialization with full mean field game refinement, and numerical experiments show stable recovery under noise and across multichannel scene families. Taken together, these ingredients establish a foundational framework for a class of inverse problems in which data are generated by optimizing and interacting populations rather than by passive signals. The framework identifies the forward MFG model, admissible observation channels, structural mechanisms, and computational recovery route needed to study active tomography under collective dynamics. Such problems arise, among other settings, in biological transport, vascular flow, and subsurface groundwater dynamics.

math.OC↗

arXiv2026

UIKA: Fast Universal Head Avatar from Pose-Free Images

We present UIKA, a feed-forward animatable Gaussian head model from an arbitrary number of pose-free inputs, including a single image, multi-view captures, and smartphone-captured videos. Unlike the traditional avatar method, which requires a studio-level multi-view capture system and reconstructs a human-specific model through a long-time optimization process, we rethink the task through the lenses of model representation, network design, and data preparation. First, we introduce a UV-guided avatar modeling strategy, in which each input image is associated with a pixel-wise facial correspondence estimation. Such correspondence estimation allows us to reproject each valid pixel color from screen space to UV space, which is independent of camera pose and character expression. Furthermore, we design learnable UV tokens on which the attention mechanism can be applied at both the screen and UV levels. The learned UV tokens can be decoded into canonical Gaussian attributes using aggregated UV information from all input views. To train our large avatar model, we additionally prepare a large-scale, identity-rich synthetic training dataset. Our method significantly outperforms existing approaches in both monocular and multi-view settings.

cs.CV↗

arXiv2026

EasyVFX: Frequency-Driven Decoupling for Resource-Efficient VFX Generation

Generating high-fidelity visual effects (VFX) typically demands massive datasets and prohibitive computational power due to the intricate coupling of spatial textures and temporal dynamics. In this paper, we introduce EasyVFX, a resource-efficient framework that achieves realistic VFX synthesis under stringent constraints. Our core philosophy lies in frequency-domain decomposition: we observe that the complexity of VFX can be significantly mitigated by decoupling high-frequency components, which represent intricate spatial appearances, from low-frequency components that encapsulate global motion dynamics. This spectral disentanglement transforms a high-dimensional learning problem into manageable sub-tasks, thereby lowering the optimization barrier and reducing data dependency. Building upon this insight, we propose a two-stage training paradigm. First, we design a Frequency-aware Mixture-of-Experts (Freq-MoE) architecture. By utilizing a soft routing mechanism, our model assigns specialized experts to distinct spectral bands, enabling them to cultivate robust priors for appearance and motion dynamics. This specialization allows the model to acquire foundational VFX knowledge with fewer GPU resources. Second, we introduce a Test-Time Training strategy powered by a novel Frequency-constraint Loss. This allows the pre-trained model to swiftly adapt to specific, unseen effects through localized optimizations, requiring only about 100 steps on a single GPU. Experimental results demonstrate that EasyVFX produces structurally consistent and visually stunning effects, proving that frequency-aware learning is a key catalyst for democratizing professional-grade VFX.

cs.CV↗

arXiv2026

The Scaling Laws of Skills in LLM Agent Systems

As agent systems scale, skills accumulate into large reusable libraries, yet their scaling laws remain poorly understood. Across 15 frontier LLMs, 1,141 real-world skills, and over 3M routing or execution decisions, we identify two coupled laws. Routing law: single-step routing accuracy decays logarithmically with library size ($R^2{>}0.97$ for all models), with errors progressing from local skill competition to cross-family drift and capture by overly general "black-hole skills". Execution law: before state realization, joint routing is approximately multiplicative, whereas correct execution can improve difficult downstream decisions by about $4{\times}$. A single parameter, the routing logarithmic decay slope $b$, couples the two laws: routing-side fits predict execution-side rescue across models, showing that the same library property controls both pre-execution collapse and downstream recoverability. The laws are actionable: law-guided optimization raises held-out routing accuracy from 71.3% to 91.7%, reduces hijack from 22.4% to 4.1%, and transfers directionally to downstream ClawBench and ClawMark execution settings, improving mean pass rate from 49.3% to 61.6% on ClawBench and from 28.4% to 34.5% on ClawMark. These results show that agent performance depends not only on model capability, but also on the structure, granularity, and exposure policy of the skill library.

cs.CL↗

arXiv2026

AcademiClaw: When Students Set Challenges for AI Agents

Benchmarks within the OpenClaw ecosystem have thus far evaluated exclusively assistant-level tasks, leaving the academic-level capabilities of OpenClaw largely unexamined. We introduce AcademiClaw, a bilingual benchmark of 80 complex, long-horizon tasks sourced directly from university students' real academic workflows -- homework, research projects, competitions, and personal projects -- that they found current AI agents unable to solve effectively. Curated from 230 student-submitted candidates through rigorous expert review, the final task set spans 25+ professional domains, ranging from olympiad-level mathematics and linguistics problems to GPU-intensive reinforcement learning and full-stack system debugging, with 16 tasks requiring CUDA GPU execution. Each task executes in an isolated Docker sandbox and is scored on task completion by multi-dimensional rubrics combining six complementary techniques, with an independent five-category safety audit providing additional behavioral analysis. Experiments on six frontier models show that even the best achieves only a 55\% pass rate. Further analysis uncovers sharp capability boundaries across task domains, divergent behavioral strategies among models, and a disconnect between token consumption and output quality, providing fine-grained diagnostic signals beyond what aggregate metrics reveal. We hope that AcademiClaw and its open-sourced data and code can serve as a useful resource for the OpenClaw community, driving progress toward agents that are more capable and versatile across the full breadth of real-world academic demands. All data and code are available at https://github.com/GAIR-NLP/AcademiClaw.

cs.AI↗