arXiv subjects

Hua Zhou

Publications and source records attributed to Hua Zhou.

At least 19 recordsLinked to original sources

arXiv2026

Evaluating Time-Series Foundation Models and Multimodal Dietary Context for CGM Forecasting

Continuous glucose monitoring (CGM) provides high-frequency measurements of glucose dynamics and enables short-term glucose forecasting for diabetes management. Although time-series foundation models have shown strong general forecasting ability, their effectiveness for CGM prediction and the added value of multimodal dietary context remain unclear. We conduct a comprehensive empirical study using eight public CGM datasets spanning Type 1 diabetes, Type 2 diabetes, and non-diabetes populations. Under a unified protocol across multiple context lengths and prediction horizons, zero-shot foundation models did not consistently outperform strong task-specific baselines such as Elastic Net and PatchTST. In contrast, lightweight fine-tuning substantially improved forecasting performance. For example, fine-tuned Chronos-Bolt reduced RMSE by 6.5%-18.4% in the T1D cohort and by 8.6%-18.2% in the non-diabetes/T2D cohort, with comparable improvements in both in-distribution and out-of-distribution test settings. We further evaluate multimodal dietary context using CGMacros, which provides temporally aligned CGM signals, food images, and macronutrient records. A residual-based fusion framework reduced overall RMSE by approximately 3% and postprandial RMSE by approximately 15% relative to the CGM-only baseline. Moreover, Chronos-based CGM representations were more strongly correlated with observed postprandial glucose increments than representations from LSTM and CatBoost, even after those models incorporated additional dietary modalities, suggesting that pretrained temporal representations better preserve meal-induced excursion patterns. These findings show that foundation models require CGM-specific adaptation for reliable forecasting and that dietary context provides clinically meaningful signals beyond CGM alone, especially during postprandial periods.

stat.ML↗

arXiv2026

Remote epitaxy beyond polarity

Remote epitaxy through a monolayer two-dimensional material-covered substrate establishes a crystallographic registry across the van der Waals (vdW) surface that enables the epitaxial growth, lift-off and transfer of single-crystalline films. A central belief in remote epitaxy is that the substrate facilitating the phenomenon must be a material with strong ionicity, as the interatomic electrostatic potential fluctuation in covalent and metallic materials is substantially attenuated by two-dimensional materials. Here, we show remote epitaxy is possible when the substrate is a metallic or covalently bonded material and experimentally demonstrate non-polar remote homo- and heteroepitaxy across a wide range of material systems, including both metals and semiconductors. The achieved non-polar remote interactions are designed and engineered by harnessing substrate conductivity and vicinal surface step-edge density. These findings indicate that remote epitaxy is universal and applicable to ionic, metallic, and covalent materials, expanding its capabilities and stimulating a plethora of new fundamental scientific questions about the mechanism of remote epitaxy.

cond-mat.mtrl-sci↗

arXiv2026

TuringLLM: Efficiently Scaling Foundation Models Toward Physical AI

We present Turing-20B-A2B, a 20B-parameter Mixture-of-Experts language model that activates approximately 2B parameters per token, designed for long-context and latency-sensitive physical AI applications. The model adopts Quantile Routing in a dynamic top-k configuration, enabling token-adaptive expert allocation while maintaining balanced expert utilization and a controlled average compute budget. During deployment, we further apply capacity-constrained routing to prompt prefill for more regular and efficient expert execution, while retaining dropless routing during pretraining. Turing-20B-A2B also employs a hybrid attention architecture that combines Lightning Attention with a small number of full-attention layers for efficient long-context modeling. The model is pretrained with a progressive three-stage curriculum and extended to a native context length of 128K through continued pretraining, with further inference-time extension to 512K using YaRN. Despite its compact active-parameter budget, Turing-20B-A2B achieves, at the base-model stage, overall general capability exceeding Qwen3-8B Base and approaching Qwen3.5-9B Base, while maintaining strong long-context performance and favorable prefill-latency scaling. These results demonstrate an effective balance among model capability, long-context scalability, and practical inference efficiency.

cs.AI↗

arXiv2026

Charge transfer and competing symmetry breaking drive orbital reconstruction and emergent ferromagnetism in insulating oxide superlattices

Electron correlation, hopping, and ligand-to-metal charge transfer collectively lead to diverse electronic and magnetic phenomena in 3$d$ transition-metal oxides, where directional d orbitals make hopping highly sensitive to symmetry-dependent orbital overlap. Heterostructure engineering with atomically flat interfaces adds symmetry-breaking charge transfer as a further route to emergent behavior, yet whether interfacial mismatch between constituent oxides of a superlattice shapes ground states independent of epitaxial strain remains unresolved. Here we examine superlattices combining NdNiO$_3$ with Mott-insulating NdMnO$_3$. Varying layer thickness and combining transport with X-ray spectroscopy, we show that electron transfer from NdMnO$_3$ to NdNiO$_3$ drives a room-temperature insulating state with a distinct electronic structure, accompanied by a reversal in orbital symmetry beyond simple strain considerations, underscoring the interface's central role. These reconstructions stabilize an emergent ferromagnetic insulating state arising from interfacial Ni$^{2+}$-O-Mn$^{4+}$ superexchange. Our results establish a pathway to interface-engineered ferromagnetic insulating phases via competing interactions, with potential for spin-insulatronic applications.

cond-mat.mtrl-sci↗

arXiv2026

Inject, Align, Recover: Staged Post-Training for Retrieval-Free Document Knowledge Internalization

Large language models often fail to answer questions about a bounded document collection when the source documents are not retrieved at inference time. We study this setting as document knowledge internalization: converting a fixed corpus into usable parametric knowledge for retrieval-free question answering. We propose IAR (Inject, Align, and Recover), a three-stage post-training framework that separates structured document knowledge injection, QA behavior alignment, and general ability recovery. Unlike conventional continued pretraining, Inject converts source documents into continuation, rewrite, and instruction-conditioned reconstruction objectives. Align then adapts the injected model with answer-only QA supervision, while Recover merges the domain-adapted model with the base instruction model to recover general capabilities. Across Common Corpus (CC) and CCI, and across Llama, Phi, Qwen, and SmolLM model families, IAR improves the domain-primary domain-general frontier for retrieval-free document internalization. In the main comparison, IAR improves over Vanilla SFT on all four reported metrics in 7 of 8 dataset-model settings, with average gains of 3.6 percentage points in domain QA accuracy and 12.1 percentage points in mean general performance across IFEval, MMLU, and MSBench. Extended CC baselines show that LoRA and FAPM can win individual general metrics, but among methods that also reach leading or near-leading domain internalization, IAR retains one of the strongest general profiles.

cs.CL↗

arXiv2026

Debiased Machine Learning for Partially Linear Accelerated Failure Time Models

The Cox model remains the default for survival analysis, but the proportional hazards assumption is often violated and hazard ratios can be difficult to interpret. Accelerated failure time (AFT) models provide an intuitive time-scale alternative, yet flexible covariate adjustment while preserving valid inference on a target exposure remains challenging. For the partially linear AFT model under right censoring, a rank-based debiased machine learning (DML) framework remains undeveloped: the rank-based pairwise moment is not Neyman orthogonal and standard cross-fitting does not directly apply to U-statistics. We develop the first such framework by combining an orthogonalized rank-based U-statistic, a censoring-corrected influence function, and block-pairwise cross-fitting, yielding valid inference under flexible nuisance estimation. Simulations and an application to All of Us electronic health record data demonstrate finite-sample performance and practical utility.

stat.ME↗

arXiv2026

Wnuan: Staged Post-Training for Question Answering over Proprietary Enterprise Knowledge

Enterprise question answering requires models to acquire proprietary knowledge without discarding general capabilities. We present Wnuan, a three-stage pipeline that constructs task-oriented supervision from documents, performs supervised fine-tuning with general-data replay, and applies reinforcement learning to residual errors. On the 707-question WnuanBench, the primary 32B route raises acceptable-answer rate (AAR) from 52.76% before adaptation to 80.06% after SFT and 91.51% after RL. Under a matched 100-update protocol, residual-error sampling outperforms full-pool and size-matched random sampling by 3.11 and 2.97 points, respectively. Source-cluster bootstrap intervals remain above zero for both contrasts, and a same-domain validation set preserves the ordering. The general-benchmark average decreases by 5.17 points across the route, concentrated in instruction following. The automatic evaluation ensemble agrees with an authoritative domain expert on 90.5% of a stratified Wnuan-Inst response sample. These results characterize both the gains and the general-capability cost of staged enterprise adaptation.

cs.AI↗

arXiv2026

TuringViT: Making SOTA Vision Transformers Accessible to All

Modern VLMs and VLA systems commonly adopt off-the-shelf ViTs such as SigLIP2 as visual encoders, but diverse downstream requirements in latency, temporal modeling, and VLM integration often call for customized SOTA-level ViTs. Training such encoders remains beyond the reach of much of the community, as it requires massive image-text data, while standard softmax attention makes high-resolution or dynamic-resolution pretraining prohibitively costly and often forces low-resolution pretraining followed by post-hoc adaptation. TuringViT addresses these challenges with three key designs: Turing Linear Attention (TLA) for efficient sequence modeling, VISTA-Curation to construct supervision-rich image-video training data, and native dynamic-resolution pretraining that supports flexible inputs from the start and transfers seamlessly to downstream VLMs. As a result, TuringViT outperforms leading open-source ViT baselines with only 10% of the data, achieves stronger downstream VLM performance, and delivers substantially better latency scaling on high-resolution inputs. Our scaling-law analysis further shows that TuringViT continues to improve predictably with curated data scale, far from saturation. Its fast adaptation, hardware-friendly design, and efficient deployment have made it a unified visual foundation across XPeng's AI systems. More broadly, TuringViT provides a reproducible pipeline that dramatically lowers the cost for the community to train, customize, and deploy SOTA-level ViTs, moving toward making such Vision Transformers accessible to all.

cs.CV↗

arXiv2026

NNLO QCD Corrections to $D$-Wave Spin-Singlet Heavy Quarkonia Decay $\eta_{Q2}\to\gamma\gamma$ via the Principle of Maximum Conformality

In this paper, we perform a comprehensive study of the decay process $\eta_{Q2}\to\gamma\gamma$ for $D$-wave spin-singlet heavy quarkonia up to next-to-next-to-leading-order (NNLO) QCD corrections within the nonrelativistic QCD effective theory. Following its factorization formalism, the total decay width is decomposed into perturbatively calculable short-distance coefficients (SDCs) and nonperturbative $D$-wave long-distance matrix elements (LDMEs). The original NNLO series of SDCs suffers from sizable renormalization and factorization scale uncertainties. To eliminate such inherent scale ambiguities, we adopt the Principle of Maximum Conformality (PMC). We show that recursively applying the renormalization group equations for the running of $\alpha_s$ and $D$-wave LDMEs within the PMC framework yields an effective strong coupling $\alpha_s(Q_\ast)$ consistent with the expansion coefficients, resulting in a scale-invariant perturbative series. The determined PMC scales are $Q_\ast=1.483$ GeV for $\eta_{c2}$ and $Q_\ast=4.246$ GeV for $\eta_{b2}$. By removing divergent renormalon contributions, the PMC naturally improves the convergence of the perturbative series for SDCs. Our PMC predictions for the total decay widths are $\Gamma_{\eta_{c2}\to\gamma\gamma}^{\rm PMC} = 3.322^{+0.899}_{-0.828}\ \text{eV}$ and $\Gamma_{\eta_{b2}\to\gamma\gamma}^{\rm PMC} = 0.0188^{+0.0014}_{-0.0013}\ \text{eV}$. The uncertainties arise from variations of the charm and bottom quark masses $\Delta m_c=\pm 0.07$ GeV, $\Delta m_b=\pm 0.06$ GeV, as well as systematic errors from uncalculated higher-order corrections. The corresponding branching ratios are $\text{Br}(\eta_{c2}\to\gamma\gamma) = \big(7.463^{+2.020}_{-1.860}\big)\times 10^{-6}$ and $\text{Br}(\eta_{b2}\to\gamma\gamma) = \big(6.460^{+0.481}_{-0.447}\big)\times 10^{-7}$.

hep-ph↗

arXiv2026

ChartWalker: Benchmarking the Cross-Chart RAG Task with Hierarchical Knowledge Graphs

Cross-Chart Retrieval-Augmented Generation (RAG) is critical for complex multi-modal analytical tasks in scientific, business, and political domains. However, existing benchmarks either focus on tables, which are well-structured and textualized, or generate cross-chart questions by simply extracting key points, which often induces lexical overlap between queries and evidence and yields logically inconsistent reasoning chains. To address this, we introduce ChartWalker, a novel framework for constructing challenging cross-chart RAG tasks. ChartWalker features a hierarchical knowledge graph construction method tailored to charts, which organizes entities and relations by granularity to preserve analytical structure. We then propose a structure-aware sampling algorithm that synthesizes semantically coherent, multi-hop reasoning paths, enabling explicit control over query difficulty and granularity for QA generation. Built with this framework, we release ChartWalker-Bench, a comprehensive benchmark spanning diverse domains and cross-chart query types. Extensive evaluations across major RAG paradigms reveal significant performance gaps, underscoring the benchmark's difficulty and utility. Furthermore, we provide ChartWalker-Agent, an agentic baseline to facilitate analysis and inspire future system design.

cs.IR↗

arXiv2026

Persistent structural distortions and absent superconductivity in trilayer nickelate thin films

A new family of high-temperature superconductors was recently discovered in the $n=2,3$ Ruddlesden-Popper nickelates, where superconductivity emerges concomitant with suppression of parent density waves and structural octahedral rotations under hydrostatic pressure. Intriguingly, compressive strain mimics the structural effects of pressure in the $n=2$ phase, yielding ambient-pressure superconductivity. However, analogous strain-stabilized superconductivity has not been realized in the $n=3$. Here, we use atomically-precise synthesis, transport, picoscale electron microscopy, and synchrotron X-ray diffraction to probe $n=3$ La$_4$Ni$_3$O$_{10}$ thin films. Although compressive strain suppresses density wave order, we do not observe superconductivity even under the largest strain state. Importantly, we identify a structural distortion unique to strained $n=3$ thin films that may inhibit superconductivity: persistent, layer-inequivalent octahedral rotations around the $c$-axis. Our results highlight key differences between the $n=3$ and $n=2$ systems, suggesting that ambient-pressure superconductivity in the $n=3$ may require new methods beyond epitaxial strain engineering.

cond-mat.mtrl-sci↗

arXiv2026

MechVQA: Benchmarking and Enhancing Multimodal LLMs on Comprehensive Mechanical Drawing Understanding

Multimodal Large Language Models (MLLMs) have demonstrated significant achievements in general visual question answering (VQA) tasks. However, they remain brittle on mechanical engineering drawings, where high annotation density and weak domain knowledge, compounded by unreliable spatial relation reasoning under strict projection rules and geometric constraints, make decisive cues easy to miss and frequently lead to wrong answers. To bridge this gap, we introduce the first comprehensive mechanical drawing understanding dataset, MechVQA, created through a semi-automated construction and quality-control pipeline. MechVQA contains 3.3k high-density pictures with 21K question-answer pairs, spanning 10 different fine-grained tasks across three capability levels: Recognition, Reasoning, and Judging, providing a testbed to evaluate and improve MLLM understanding on real-world mechanical drawings. On top of MechVQA, we then develop the MechVL model through a multi-stage training paradigm, building a strong domain-specialized baseline. Extensive experimental results demonstrate that MechVL outperforms the strongest closed-source baseline by 7.57 percentage points on the MechVQA total score, significantly enhancing mechanical drawing understanding ability and providing a reusable foundation for deploying MLLMs in mechanical design and inspection scenarios.

cs.CV↗

arXiv2026

RAFT: Data Refinement and Adaptive Distillation for Domain Fine-Tuning with Alleviated Forgetting

Domain-specific supervised fine-tuning (SFT) often improves in-domain performance at the cost of degrading a model's general capabilities. We view this degradation through two practical gaps in domain SFT: a supervision-compatibility gap, where domain targets differ in style and reasoning format from the original model's natural responses, and a trajectory-preservation gap, where teacher-forced SFT optimizes fixed target tokens without constraining the model's behavior on its own generated prefixes. This process fails to preserve the model's original behavior. We propose RAFT (Data Refinement and Adaptive Distillation for Domain Fine-Tuning with Alleviated Forgetting), a two-stage framework that addresses both factors. First, RAFT constructs model-compatible supervision through self-conditioned rewriting, semantic filtering, and answer fusion. Second, RAFT performs Answer-Conditioned On-Policy Distillation, where the original instruction-tuned model provides soft targets on student-generated trajectories while being conditioned on the fused answer as helpful context. We further introduce top-K temperature distillation and EMA-based adaptive loss balancing to stabilize the domain-general trade-off. Across three instruction-tuned backbones and five domains, RAFT improves average domain accuracy by 23.2% over standard SFT, while recovering part of the SFT-induced degradation on MS-Bench and IFEval, with relative improvements of 18.2% and 10.2%, respectively. These results show that coupling data refinement with trajectory-level preservation provides an effective recipe for domain fine-tuning with alleviated forgetting.

cs.LG↗

arXiv2026

User-Centric Clustering for uRLLC in Cell-Free RAN via Extreme Value Theory

Ultra-reliable low-latency communication (uRLLC) is a pivotal enabler for B5G/6G networks, yet it faces severe challenges from rare but critical extreme events, which are characterized by heavy tails in the delay distribution. While the cell-free radio access network (CF-RAN) architecture offers essential spatial diversity to combat these uncertainties, conventional user-centric clustering designs typically focus on average metrics, thereby inadequately addressing such tail behaviors. We propose a novel, tail-risk-aware, user-centric clustering framework operating within the finite blocklength (FBL) regime. Our approach employs extreme value theory (EVT), specifically the peaks-over-threshold (POT) model, to accurately quantify the probability of queue latency violations. This framework is applied to formulate an energy efficiency (EE) maximization problem under strict tail latency constraints. The problem is solved via an efficient online algorithm that integrates Lyapunov optimization with successive convex approximation (SCA). Simulation results demonstrate that the proposed scheme, through its dynamic adaptation of cluster formation to mitigate tail risks, achieves a superior reliability-efficiency trade-off and leads to a significant suppression of extreme latency events.

cs.IT↗

arXiv2026

Deterministic fabrication of large-area, high-crystallinity oxide moire superlattices

Oxide twistronics extends moire engineering beyond van der Waals materials, offering a promising platform for accessing emergent interfacial phenomena arising from the strong coupling of lattice, charge, and orbital degrees of freedom in complex oxides. However, deterministic fabrication of high-crystallinity oxide moire superlattices over large lateral dimensions remains challenging due to the three-dimensional bonding network of oxides. Here, we demonstrate a scalable, generalized fabrication strategy that enables the formation of high-crystallinity oxide moire superlattices with clean, chemically bonded interfaces and precisely controlled twist angles down to nominal values of 0.1 degree, achieving sub-degree twist-angle accuracy across large contiguous lateral dimensions approaching the millimeter scale. Using NaNbO3 as a model system, we show that the resulting interlayer coupling drives pronounced structural reconstruction that modifies both the phase structure and ferroelectric domain configuration. Synchrotron-based X-ray 3D reciprocal space mapping reveals the emergence of a single-phase state in twisted bilayers, in contrast to the mixed-phase structure observed in single-layer membranes prior to twist assembly. The structural signatures are further consistent with gradual lattice rotation distributed along the thickness direction that may accommodate interfacial shear strain, distinct from reconstruction observed in van der Waals moire systems, which primarily occurs through in-plane stacking rearrangement. This collective lattice response is correlated with twist-dependent nanoscale electromechanical modulations observed by piezoresponse force microscopy. These results establish a scalable materials platform for oxide twistronics and open new pathways towards integrating twist-engineered complex oxides into practical, macroscale device architectures.

cond-mat.mtrl-sci↗

arXiv2026

Tactics for Improving Least Squares Estimation

This paper deals with tactics for fast computation in least squares regression in high dimensions. These tactics include: (a) the majorization-minimization (MM) principle, (b) smoothing by Moreau envelopes, and (c) the proximal distance principle for constrained estimation. In iteratively reweighted least squares, the MM principle can create a surrogate function that trades case weights for adjusted responses. Reduction to ordinary least squares then permits the reuse of the Gram matrix and its Cholesky decomposition across iterations. This tactic is pertinent to estimation in L2E regression and generalized linear models. For problems such as quantile regression, non-smooth terms of an objective function can be replaced by their Moreau envelope approximations and majorized by spherical quadratics. Finally, penalized regression with distance-to-set penalties also benefits from this perspective. Our numerical experiments validate the speed and utility of deweighting and Moreau envelope approximations. Julia software implementing these experiments is available on our web page.

stat.CO↗

arXiv2026

New Determinations of the Charm and Bottom Quark Masses Using QCD Quarkonium Sum Rules

We reanalyze the perturbative QCD (pQCD) corrections to quarkonium QCD sum rules and extract the heavy quark masses $\overline{m}_{q}(\overline{m}_{q})$ ($q=c,b$). At present, the pQCD corrections to the correlation functions of two heavy-quark pseudoscalar and vector currents at zero momentum transfer, denoted as $M_{n,q}^{X,\rm th}$ ($X = P, V$), are calculated up to the $\mathcal{O}(α_s^3)$ order. These corrections exhibit significant renormalization scheme and scale dependence, which introduces large theoretical uncertainties and deteriorates the precision of heavy quark mass determinations. In this work, we eliminate the renormalization scheme and scale ambiguities in the perturbative part of $M_{n,q}^{X,\rm th}$ by adopting the Principle of Maximum Conformality (PMC) within the characteristic operator (CO) approach. The CO approach, a novel extension of the standard PMC procedure, simultaneously determines the effective coupling $α_s(Q_*)$ and the effective quark mass $\overline{m}_q(Q_*)$. It systematically absorbs the nonconformal $\{β_i\}$-terms and $\{γ_i\}$-terms via the renormalization group equations, yielding a strictly scheme- and scale-independent conformal perturbative series. Based on the improved PMC conformal series, we further provide reliable estimates for the unknown $\mathrm{N^4LO}$ contributions using the Padé approximation method. The final predicted heavy quark masses in the $\overline{\mathrm{MS}}$ scheme read: $\overline{m}_c(\overline{m}_c)=1275.8\pm 0.4~\text{MeV}$, extracted from the second moment of the charmed pseudoscalar correlator $M_{2,c}^{P}$; and $\overline{m}_b(\overline{m}_b) = 4177.0 \pm 7.2~\text{MeV}$, extracted from the first moment of the bottom vector correlator $M_{1,b}^{V}$. Both results agree well with the PDG world averages with deviations smaller than $1σ$.

hep-ph↗

arXiv2026

Gravitational Waves from Higgs Preheating after Inflaton $Z_2$-Symmetry Breaking

In this paper, nonperturbative lattice simulations are used to study Higgs preheating and the associated gravitational wave (GW) background after the inflaton $Z_2$ symmetry is broken during inflation. This symmetry breaking generates both trilinear and quartic inflaton-Higgs interactions during preheating. The quartic inflaton-Higgs coupling is characterized by $q_{ϕh}\equiv λ_{ϕh}/λ_ϕ$, while the trilinear interaction enters jointly through $q_{ϕh}$ and $q_ε\equiv m_ϕ/(\sqrt{λ_ϕ}ϕ_0)$. The Higgs self-coupling parameter $q_h\equiv λ_h/λ_ϕ$ determines the onset of backreaction through the effective mass induced by Higgs self-interactions. Our simulations show that efficient preheating requires both a sufficiently broad resonance band and delayed backreaction. For $λ_ϕ=10^{-13}$, the viable parameter region is approximately $10<q_{ϕh}<10^4$, $q_h<10^3$, and $q_ε<10^{-5}$. Smaller $q_ε$ keeps the system in a quartic-dominated regime and suppresses the rapid drift of resonance bands, while smaller $q_h$ delays the end of preheating by weakening self-interaction-induced backreaction. The amplified Higgs inhomogeneities source GW through the transverse-traceless part of the anisotropic stress tensor. The lattice results show that the GW spectrum grows rapidly during parametric resonance, broadens through rescattering, and saturates in the nonlinear stage. At late times, the spectrum develops a broad peak with amplitude $Ω_{\rm gw}\sim10^{-6}$ at production. After redshifting to the present day, the peak frequency is $f\sim10^9\,{\rm Hz}$ with present-day amplitude $Ω_{\rm gw,0} h^2 \sim 10^{-10}$. These results suggest that high-frequency GW from Higgs preheating may be detectable by future resonant-cavity detectors.

hep-ph↗