Searcharxiv⌕ Search

arXiv subjects

Jie Zhu

Publications and source records attributed to Jie Zhu.

At least 19 recordsLinked to original sources
arXiv2026

MemX: Memory-Aware Automatic Parallelism for Cross-Vendor Heterogeneous GPU Training

Cross-vendor heterogeneous GPU clusters are increasingly used for large-scale model training. However, their GPUs differ substantially in compute power, memory capacity, and interconnect characteristics. Existing automatic parallelism training systems usually minimize iteration time while treating memory as a hard constraint, almost failing in resolving the compute--memory asymmetry of cross-vendor clusters regardless correlation between memory and throughput. In this paper, we present MemX, a new memory-aware automatic parallelism system that jointly optimizes iteration time, peak memory usage, and cross-GPU memory balance. MemX consists of a weighted-product objective function, heterogeneous recomputation strategies, a minimal-set profiler, interpolation-based data augmentation, a dual-output XGBoost predictor, and a hash-signature prediction cache, in order to accelerate search of large-scale auto-parallel best configurations at lower cost. We implement MemX on Megatron-LM and evaluate it against four automatic parallelism baselines on two heterogeneous clusters. MemX achieves the lowest prediction Mean Absolute Percentage Error (MAPE) of 12.17% and is the only method that identifies the real Top-1 configuration under both time-based and memory-balance rankings. In terms of search cost and search space, MemX searches 31,240 configurations in 26.5 seconds, while an end-to-end profiling baseline is extrapolated to require about 17 days at the same scale. The simulation study that extends the measured 20-GPU data to 40-GPU and 60-GPU clusters further illustrates the effectiveness of MemX at larger scales.

cs.DC↗
arXiv2026

Phase-Coordinated Model Predictive Control for AI Training Power Smoothing

Periodic power fluctuations from artificial intelligence (AI) training workloads can excite oscillations in power systems. This paper proposes a model predictive control (MPC) framework that exploits short start-time delays to mitigate power fluctuations at the computing source. By coordinating the relative phases of independent training jobs, the proposed method reduces the superposition of oscillatory power components without interrupting running jobs or requiring additional power-smoothing hardware. To obtain a tractable optimization formulation, cosine and sine projections are employed to quantify the dominant oscillatory power components, and their magnitudes are used to construct a convex quadratic power-smoothing objective. The receding-horizon framework updates launch decisions as new workload information becomes available while maintaining scheduling feasibility under limited computation time. Numerical results demonstrate reductions of 74.39\% in mean-square power fluctuations within the 0.2--2-Hz band and 46.6\% in power standard deviation, with an average launch delay of 2.188~s. All launch deadlines are satisfied, and the maximum end-to-end computation time is 0.390~s, below the 0.50-s MPC update period.

eess.SY↗
arXiv2026

Unveiling the Secret of AdaLN-Zero in Diffusion Transformer

Diffusion transformer (DiT), a rapidly emerging architecture for image generation, has gained much attention. However, despite ongoing efforts to improve its performance, the understanding of DiT remains superficial. In this work, we delve into and investigate a critical conditioning mechanism within DiT, adaLN-Zero, which achieves superior performance compared to adaLN. Our work studies three potential elements driving this performance, including an SE-like structure, zero-initialization, and a "gradual" update order, among which zero-initialization is proved to be the most influential. Building on this understanding, we propose an analysis-guided initialization strategy, termed adaLN-Gaussian, which serves both as an empirical validation of our analysis and as a practical initialization method that consistently improves optimization efficiency. On the other hand, inspired by the SE-like structure, we introduce an improved conditioning mechanism called SE-adaLN-Zero. Extensive experiments following DiT on four datasets, especially on ImageNet1K, demonstrate the effectiveness and generalization of adaLN-Gaussian and SE-adaLN-Zero. Beyond class-to-image generation, we also evaluate the generalization of the two improved methods on text-to-image generation.

cs.CV↗
arXiv2026

Natural Code Transformation Confuses Agents in Vulnerability Analysis

Having demonstrated a strong potential for automating vulnerability discovery, AI agents can also be used to attack software systems at scale. Today's software systems need mechanisms to level the playing field by slowing down potential attacks, giving them time to patch vulnerabilities and strengthen software. Along this line, our study of agent trajectories in vulnerability analysis shows a surprising finding: agents produce dramatically different results across real-world code commits that carry the same functionality and the same vulnerability. This phenomenon appears consistently in multiple code domains and vulnerability categories. It motivates us to study natural and functionality-preserving code transformations derived from real-world software development, and their potential as a defense against attacker agents. We build a corpus of natural transformations by mining recurring development patterns from real-world code commits, and apply them to evaluate impact on agentic vulnerability analysis. Our results show that even a single form of transformation can effectively "confuse" frontier agents, e.g., Codex with GPT-6 Sol, reducing successes in vulnerability discovery. We also study different agent failures, whether transformations can be composed to provide stronger protection, and adaptive attackers who seek to reverse them. Overall, our study of natural code transformation significantly expands the research space of using code transformations to defend against malicious agents.

cs.CR↗
arXiv2026

Single-step Retrosynthesis via Reinforcement Learning from Forward Feedback

Single-step retrosynthesis is a one-to-many inverse problem, yet most models are trained and evaluated against a single recorded reactant set. Reference-based training gives no explicit positive signal for unrecorded alternatives, and exact-match evaluation rejects non-matching predictions regardless of chemical plausibility. We propose Retrosynthesis via Reinforcement Learning from Forward Feedback (RetroRLFF), which optimizes reactant generation using both reference agreement and forward consistency with the target product. Starting from a supervised decoder-only Transformer, RetroRLFF generates complete reactant candidates and uses a frozen forward reaction model to predict products. A sequence-level reward combines reference agreement and forward consistency while penalizing invalid outputs. Direct Advantage-Based Policy Optimization updates the policy without human preference annotations or a preference-trained reward model. On USPTO-50K, RetroRLFF outperforms all 14 baselines in candidate-level round-trip accuracy at k = 1, 3, 5, and 10. With approximately 50 million parameters, it achieves 90.1% top-1 round-trip accuracy, exceeding R-SMILES, the strongest competing baseline on this metric, by 8.5 percentage points. Compared with its supervised policy, round-trip accuracy increases from 83.1% to 90.1%, while exact-match accuracy is maintained at 57.9% versus 57.5%. In an anonymized expert assessment of 131 randomly selected test cases, organic synthesis experts prefer RetroRLFF over EditRetro in 67.9% of comparisons. These results support forward feedback as a training signal for improving forward consistency while retaining top-1 reference recovery. Expert assessment provides complementary evidence of synthetic feasibility on the selected cases, although neither forward consistency nor expert preference establishes experimental feasibility.

cs.AI↗
arXiv2026

Subtoken Vision Transformer for Fine-grained Recognition

We present Subtoken Vision Transformer (SubViT), a selective image tokenization method for fine-grained visual recognition. Standard Vision Transformers compress each fixed-size patch into a single token, although fine-grained distinctions often depend on localized variations within only a few patches. SubViT addresses this mismatch by representing discriminative patches with multiple subtokens while retaining the original token sequence for global context, thereby allocating additional capacity where it is most needed. Since attention heads encode complementary semantics and extracting attention maps at inference requires an extra backbone forward, we adopt a two-stage training strategy. Stage 1 fine-tunes the ViT using subdivision regions sampled from random attention heads, exposing the model to diverse subdivision patterns. Stage 2 identifies informative attention maps through feature-degradation distances and distills them into a lightweight single-map router, which directly predicts deterministic token-importance scores without a separate attention forward. We evaluate SubViT on Generalized Category Discovery (GCD), a challenging task requiring both fine-grained discrimination and generalization to unlabeled novel categories. Across CUB, FGVC-Aircraft, and Stanford-Cars, SubViT improves the average novel-category accuracy of DINOv2 from $81.3\%$ to $84.7\%$, with only $0.50$ ms additional latency and $3.4\%$ more FLOPs, while reducing latency by $73.8\%$ relative to Retina Patch. Code: \href{https://github.com/jiezhu23/SubViT_ACCV26}{SubViT}.

cs.CV↗
arXiv2026

Multiplexed chiral and helical acoustic anomaly bulk states in an inverse-designed metamaterial

Valley and pseudospin are fundamental degrees of freedom (DOFs) in topological acoustics. However, they have so far been realized mostly in separate systems. Here, we achieve the multiplexing of valley and pseudospin DOFs in one acoustic system. An acoustic metamaterial simultaneously hosting single and double Dirac cones is inversely designed by the developed topology optimization method. Thereafter, under a hard-boundary condition, valley-locked chiral states and pseudospin-locked helical states emerge within two separate frequency windows and can be selected by tuning the operating frequency. Furthermore, their applications for multifrequency acoustic energy enhancement are experimentally demonstrated. Our work provides a strategy to modulate acoustic waves carrying distinct topological DOFs within a single integrated platform, facilitating the development of multiplexed acoustic devices with robustness.

physics.optics↗
arXiv2026

Semileptonic $\bar{B}^*_{c} \to (P, V) \ell^- \barν_{\ell}$ decays within the self-consistent light-front quark model

Motivated by the recent observation of the $B_c^*$ meson by the ATLAS Collaboration and its promising discovery potential at future high-luminosity colliders, we investigate the semileptonic decays $\bar{B}^*_{c} \to (P, V) \ell^- \barν_{\ell}$ ($P=D$, $B_{d,s}$, $η_c(1S,\,2S,\,3S)$; $V=D^*$, $B^*_{d,s}$, $ψ(1S,\,2S,\,3S)$) in the Standard Model. Using the relevant form factors obtained in the self-consistent covariant light-front quark model, we present theoretical predictions for several physical observables, including the branching fraction, the ratios of branching fractions $R_{(L)}$, the longitudinal polarization fraction of the daughter vector meson, the $τ$ lepton spin asymmetry and the forward-backward asymmetry. It is found that the branching fractions for the Cabibbo-favored $\bar{B}^*_{c}$ $\to$ $\bar{B}^*_{s} e^- \barν_e$ and $\bar{B}^*_{s} μ^- \barν_μ$ decays reach the order of ${\cal O}(10^{-6})$, which might be firstly observed at the future high-luminosity Large Hadron Collider.

hep-ph↗
arXiv2026

Spin-Orbit Induced Confinement of Correlated Bound States in the Continuum

Repulsively bound doublons are two-particle composites formed by strong interactions and are usually separated from the scattering continuum. Introducing spin-orbit coupling fundamentally alters the underlying band structure, providing a powerful tuning knob to shift these isolated pairs toward this continuum. However, because entering such a regime typically dictates immediate dissociation, whether this coupling can drive these pairs inside while preserving their bound nature constitutes a fundamental unresolved challenge. Here we show that spin-orbit coupling in the one-dimensional Fermi-Hubbard model can drive doublons into the two-particle scattering continuum. Most of these states hybridize with extended channels and decay, whereas a subset remains decoupled and spatially bound, forming many-body bound states in the continuum (BICs). We map the interacting two-particle problem onto a two-dimensional lattice of coupled acoustic cavities, and experimentally observe both the radiating doublon continuum and the confined BIC states. These results demonstrate that spin-orbit coupling can turn selected doublons into interaction-induced BICs, deepening the understanding of continuum physics for interaction-bound pairs.

cond-mat.str-el↗
arXiv2026

Testing the strong equivalence principle with multimessenger binary neutron star mergers

The constancy of the gravitational constant $G$ is a cornerstone of the strong equivalence principle and of general relativity, yet its possible temporal variation remains a key target in tests of fundamental physics. Gravitational-wave (GW) astronomy, especially when combined with electromagnetic observations, provides an unprecedented new opportunity to probe this principle in the strong-field and dynamical regime. In this work, we develop a GW waveform model with a slowly varying gravitational constant, incorporating its effects both on compact binary dynamics and GW propagation in an expanding universe. Applying this framework to the binary neutron star merger GW170817, together with independent electromagnetic constraints on the luminosity distance, sky localization and binary inclination from GRB 170817A, we perform a joint Bayesian analysis that disentangles varying-$G$ effects from astrophysical degeneracies. We find no evidence for a temporal variation of the gravitational constant, and constrain its fractional time derivative to $\dot{G}/G \in [-3.36 \times 10^{-9}, 5.34\times10^{-10}]~{\rm yr^{-1}}$, representing the most stringent bounds obtained to date from real GW observations. Our results demonstrate the power of multi-messenger astronomy as a precision probe of the strong equivalence principle in the relativistic regime.

gr-qc↗
arXiv2026

Generation of TeV Photons by PeV Neutrinos in Dense Astrophysical Environments

Recent observations by IceCube and KM3Net of PeV-scale ultra-high-energy (UHE) neutrinos, together with detections of TeV-PeV photons from various sources such as the Crab Nebula, the Galactic Center, and gamma-ray burst by ground-based observatories including Tibet AS$γ$, MAGIC, Carpet-3, and LHAASO, point to the existence of extreme astrophysical environments capable of accelerating particles to ultra-high energies. These findings motivate investigations of possible connections between UHE neutrinos and photons in such environments. Theoretically, dense regions surrounding compact objects can efficiently produce UHE neutrinos. In this work, we calculate the production of UHE photons from neutrino-nucleon interactions, and note that if these interactions occur in the outer, optically thin regions of dense environments, the resulting photons could potentially be observed. In our model, an incident neutrino scatters off a nucleon, generating secondary partons that hadronize into pions and subsequently decay into UHE photons. We calculate the resulting photon energy spectra and find that for incident (anti)neutrinos with energies above 1 PeV, the probability of producing photons with energies exceeding 1 TeV is greater than 13%. As a concrete application, we show that this mechanism can quantitatively account for the preburst TeV photons observed in GRB 221009A, providing a natural explanation for both their energies and lead times. These findings establish a plausible mechanism linking UHE neutrino events to gamma-ray observations, providing new insights into hadronic processes in extreme astrophysical environments and supporting multi-messenger astronomy studies.

astro-ph.HE↗
arXiv2026

Study on Thickness and Temperature Dependence of Thermoelectric Properties in SnS Nanofilms

SnS as an environmentally friendly, cost-effective, and earth-abundant narrow-bandgap semiconductor material, has demonstrated significant application potential in the field of medium-temperature thermoelectric conversion. However, the thermoelectric performance of its bulk counterpart is inherently constrained by intrinsic point defects (e.g., vacancies) and the material's specific band structure. Low-dimensional engineering has emerged as a pivotal strategy for overcoming these limitations and enhancing thermoelectric performance. In this work, we systematically investigate the thermoelectric properties of SnS nanofilms with distinct thicknesses (82 nm, 199 nm, 616 nm, and 813 nm) across a temperature range of 300-600 K. Measurements were conducted using time-domain thermoreflectance (TDTR) and a dedicated thin-film thermoelectric parameter test system (ZEM-3). Our results confirm that low-dimensionalization effectively boosts the thermoelectric performance of SnS, with the thermoelectric figure of merit (ZT) displaying a pronounced dependence on both film thickness and temperature. All four SnS thin films exhibit thermoelectric performance that is markedly superior to that of bulk SnS. This enhancement is primarily attributed to the quantum confinement effect, energy filtering effect, and intensified phonon scattering, all of which are induced by the low-dimensional structural characteristics. This work provides not only experimental evidence and theoretical insights for the performance optimization of SnS nanofilms but also establishes a foundational framework for the development of high-efficiency, eco-friendly medium-temperature thermoelectric materials, thereby holding significant scientific value and practical implications.

cond-mat.mtrl-sci↗
arXiv2026

Stochastic Model Predictive Control under AC Power-Flow Constraints Using Generative Learning

This paper proposes an end-to-end generative framework for efficiently solving multi-period and multi-scenario stochastic model predictive control (SMPC) problems under nonlinear AC power-flow constraints. Conventional deterministic neural surrogates rely on a single-shot prediction, which makes reliable feasibility difficult to achieve because one dispatch trajectory must simultaneously satisfy nonlinear constraints across all scenarios and time periods. To address this limitation, a conditional stochastic neural generator (CSNG) is developed to produce multiple candidate dispatch trajectories for each uncertainty instance, enabling the recovery of a feasible and economical solution through candidate selection. A feasibility-aware self-supervised distribution-shaping scheme is further introduced to promote constraint satisfaction, candidate diversity, and operating economy without requiring computationally expensive SMPC solution labels, while mitigating candidate collapse in generative ACOPF learning. To support efficient end-to-end training, a constraint-aware differentiable architecture is introduced. It employs a projection mechanism to exactly enforce box and ramping constraints while preserving informative gradients near active bounds, together with a differentiable equality-completion surrogate for efficient AC power-flow reconstruction. Case studies on the IEEE 14- and 118-bus systems demonstrate $100\%$ feasibility, optimality gaps below $2\%$, and computational efficiency suitable for intraday dispatch. The implementation is publicly available at https://github.com/JieZhu6/Generative_SMPC.

eess.SY↗
arXiv2026

Beyond Higher-Pulse Rectification: Operational Harmonic Coordination in Renewable P2H Systems

Thyristor rectifiers (TRs) are cost-effective electrolysis power supplies for renewable power-to-hydrogen (ReP2H) systems, but their harmonics may violate grid-code limits. In contrast to conventional solutions that rely on higher-pulse (such as 24-pulse) rectifiers, this paper proposes an operational harmonic coordination scheme that enables low-cost 12-pulse TRs to meet harmonic requirements through coordinated operation. First, a harmonic model quantifies the effects of rectifier transformer (RCT) tap positions and electrolytic currents, enabling harmonic cancellation among multiple electrolyzers (ELZs). A two-layer framework then coordinates hydrogen production and harmonic mitigation. Hourly scheduling determines ELZ commitment within the harmonic feasible region under renewable uncertainty using stochastic programming and a modified progressive hedging algorithm, while minute-level dispatch tracks renewable power and mitigates harmonics. A decomposition algorithm separates production dispatch from harmonic mitigation to improve computational efficiency. Case studies based on real-life projects show that the proposed method increases profit by 31% over current-only regulation. Annual simulations further show that coordinated 12-pulse TRs can replace 24-pulse rectifiers for harmonic compliance by exchanging additional RCT tap actions for lower transformer investment, reducing rectification-stage cost by 37.5%.

math.OC↗
arXiv2026

Layer-number parity induced topological phase transition

We demonstrate that stacking topologically trivial layers, under enforced symmetry restrictions, yields emergent topological phases with protected boundary states. Remarkably, the number of layers itself acts as a topological switch, enabling the system to host topological bound states in the continuum (BICs). Analytically, we reveal that an odd-layer configuration not only renders the spectrum gapless but also supports BICs protected by interlayer reflection symmetry. Combined with entanglement-spectrum calculations, this confirms that odd-layer systems indeed support topological BICs. We provide experimental confirmation of these topological states in stacked acoustic lattices. Our findings establish a previously overlooked pathway to topology and demonstrate a readily applicable strategy for realizing exotic states in a wide range of artificial material systems.

cond-mat.mtrl-sci↗
arXiv2026

Conformation-Mediated Kinetics of Polymer Chain Scission under Tension

Chain scission is a key molecular process underlying damage and fracture in polymer networks. In this Letter, we develop a statistical-mechanical framework for predicting chain-scission kinetics while accounting for three-dimensional (3D) conformational fluctuations. Within transition-state theory, scission is formulated as a multichannel first-rupture problem, with bond-specific rates governed primarily by self-consistent potentials of mean force. In the freely jointed limit, the additional 3D configurational freedom enhances rupture relative to the collinear 1D reference. Finite bending stiffness introduces orientational correlations that can reverse this enhancement and, at high stiffness, reduce rupture rates by orders of magnitude. These correlations also make rupture bond-position dependent, with higher rates near the chain ends and a common interior rate. For sufficiently long chains, the interior contribution dominates, yielding linear scaling of the chain-scission rate with chain length. These molecularly resolved rates provide physically grounded inputs for future network-scale models of polymer damage and fracture.

cond-mat.soft↗
arXiv2026

Empowering Polymeric Materials Discovery by Artificial Intelligence

Polymeric materials underpin modern technologies spanning energy storage, microelectronics, healthcare and sustainable manufacturing. Yet their rational design remains exceptionally challenging because material performance emerges from complex interactions among molecular composition, chain architecture, processing history and hierarchical structural evolution across multiple length and time scales. Consequently, polymer research has long relied on labor-intensive experimentation and fragmented modeling approaches, limiting both mechanistic understanding and innovation efficiency. Recent advances in data infrastructure, machine learning, large artificial intelligence (AI) models and laboratory automation are beginning to reshape this landscape. Rather than functioning as isolated tools, polymer databases, predictive models, AI agents and automated laboratories are increasingly converging into interconnected discovery ecosystems. As a result, the central challenge is shifting from improving predictive accuracy alone to enabling reliable decision-making, adaptive learning and seamless integration across computation, experimentation and scientific reasoning. We argue that polymer science is entering an era of autonomous discovery, in which data, simulation, reasoning and experimentation operate within self-improving feedback loops that continuously generate hypotheses, design materials, execute experiments and refine predictive models. By unifying molecular design, process optimization, experimental validation and industrial translation, such autonomous ecosystems establish a more predictive, reproducible and scalable paradigm for polymer innovation, fundamentally transforming how polymer research is conducted.

physics.chem-ph↗
arXiv2026

SapiensID 2.0: Aligning Human Recognition Foundation Models with Human Perception

While foundation models have significantly advanced human recognition across diverse modalities, they predominantly rely on static, geometric feature extraction. This approach fundamentally diverges from human perception. Consequently, current models often suffer from "semantic blindness," overfitting to transient noise while failing to leverage invariant soft biometrics, and struggle to capture temporal motion signatures. To bridge this gap, we propose SapiensID 2.0, a human recognition framework enriched with both semantic and temporal awareness. To overcome the lack of soft-biometric annotations, we transfer zero-shot semantic knowledge from Multimodal Large Language Models (MLLMs) into a discriminative embedding space. We resolve the dimensional mismatch between these spaces using Invariant Trait Alignment (ITA) to distill core persistent traits, and Transient Noise Disentanglement (TND) to decouple artifacts like clothing. Furthermore, we design a Kinematic Semantic Attention Head (K-SAH) that extends spatial attention across temporal windows. By tracking semantic patches over time, K-SAH captures rich kinematic signatures without requiring large-scale video datasets. Extensive experiments demonstrate that SapiensID 2.0 achieves state-of-the-art performance across image- and video-based person re-identification and gait recognition, while maintaining robust face recognition capabilities.

cs.CV↗