arXiv subjects

Jie Zhao

Publications and source records attributed to Jie Zhao.

At least 19 recordsLinked to original sources

arXiv2026

ReuseGR: Offline--Online Decoupled Generative Retrieval via Adaptive Prefix Reuse

Larger generative retrieval models can improve retrieval quality, but their decoding cost makes low-latency serving difficult. When different users submit the same query, generating complete Semantic IDs (SIDs) for every request repeats computation, while their interaction histories call for different results. The challenge is to reuse large-model computation across requests while keeping retrieval personalized. We propose \textbf{ReuseGR}, which separates query-level candidate planning from request-level personalized completion. A 7B offline planner generates SID prefix sets shared across requests with the same query. A 0.5B online model uses each request's history to select prefixes and generate the remaining tokens. ReuseCut chooses a cache depth for each query based on historical coverage, balancing computation reuse and candidate coverage. To strengthen online completion, we transfer 7B suffix-scoring knowledge to 0.5B through path-level knowledge distillation (Path KD). BudgetCut then selects scored branches to improve final top-100 recall under a fixed branch budget. Its selection policy is initialized by on-policy distillation (OPD) and refined through group-relative policy optimization (GRPO). On an industrial search-advertising dataset with approximately 2M query--item records and 3.38M catalog items, ReuseGR improves query-level HR@100 by 3.83 percentage points and reduces mean online latency by 38.9\% relative to an all-online 0.5B baseline when prefixes are prepared. It achieves HR@100 comparable to an all-online 7B baseline with 57.7\% lower online latency. A 14-day user-level online A/B test against the 0.5B baseline shows relative gains of 0.93\% in click-through rate and 1.05\% in revenue per thousand requests.

cs.IR↗

arXiv2026

Quantum Correlations Between Two Giant Dipole Resonances in Ultra-Peripheral Collisions

Mutual Coulomb dissociation in ultra-peripheral heavy-ion collisions is dominated by Giant Dipole Resonance (GDR) excitation, whose subsequent decays produce forward neutrons. Standard descriptions treat the two GDRs excited in opposite nuclei as statistically independent, leading primarily to a quadrupolar azimuthal correlation determined by the common photon-polarization axis. In this work, we investigate the possibility that the two GDRs are instead prepared with a correlated quantum phase during the overlap of the electromagnetic pulses. We develop a phase-sensitive GDR decay model and implement it in a modified version of the \texttt{$n_{0}^{0}n$} Monte Carlo generator, allowing a quantitative comparison among three competing mechanisms: the weak dipolar modulation from electromagnetic recoil, the quadrupolar correlation from independent polarized GDR decays, and the phase-sensitive dipolar modulation associated with the proposed phase-lock scenario. For neutron pairs with small relative emission-time differences, the correlated-phase scenario produces a pronounced first-harmonic modulation in the two-neutron relative-azimuth distribution. Averaging over all relative emission times substantially reduces this modulation. Within the adopted energy-dependent angular model, the time-selected dipolar coefficient increases in magnitude with pair-mean neutron kinetic energy. These results identify neutron relative-azimuth correlations and their energy dependence as experimentally accessible observables for probing quantum phase correlations between the two GDR excitations. Such phase-sensitive correlations could also provide a basis for future tests of quantum entanglement between spatially separated GDR excitations at RHIC and the LHC.

nucl-th↗

arXiv2026

One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents

Repository-level software engineering (SWE) comprises heterogeneous task categories, whose progress under pooled agentic reinforcement learning can be uneven: gains in some categories coincide with regressions in others, while aggregate resolution obscures these changes. Motivated by this category see-saw, we develop a category-aware expert-training and policy-integration framework. Executable task construction and SWE Labeler, an evidence-grounded multi-axis labeling system, organize the training pools. Initial category-specific RL improves average training success while leaving uneven instance-level progress, motivating explicit consolidation of successful behavior and policy-adaptive task selection. Same-origin category experts alternate long-horizon Agentic-miniRL with Refresh-Repair-Expand (RRE): the updated policy refreshes instance mastery, reuses its own verified successful trajectories for Repair SFT, and reselects tasks for further RL. Label-routed multi-teacher on-policy distillation (MOPD) consolidates the experts into one deployable student, with ReLU-gated reward extrapolation keeping only each teacher's improving direction over the reference. Expert training and policy integration require no external model to provide solution trajectories or action targets. On a fixed, audit-filtered Pro-618 subset, we connect Pooled RL and Balanced RL, expert development, and single-model integration through overall and per-category resolution and expert-gain recovery. In final-model evaluations, MOPD achieves mean resolution of 86.19% on SWE-bench Pro V2 and 71.78% on SWE-bench Multilingual, exceeding the base model by 4.57 and 3.89 percentage points, respectively. The same three 27B experts also improve a Qwen3.6-35B-A3B student across the evaluated task sets.

cs.SE↗

arXiv2026

Force Sensing in a Recoverable Bosonic Quantum Memory

In continuous-variable quantum information, sensing and quantum error correction in the same mode face an intrinsic trade-off: sensors must accumulate weak external displacements, whereas bosonic codes typically suppress such shifts as errors. Here, we establish compatibility by constructing finite-energy bosonic codes that enable a single oscillator to sense an external force while actively protecting an unknown logical qubit. By encoding logical amplitudes across discrete wave-packet sites and capturing the force signal in a shared internal coordinate, we achieve an endpoint probe readout that is independent of the logical state under continuous photon loss $γ$. Crucially, binomial site recovery cancels packet-dependent probe distortions alongside logical errors: for a $(2r+2)$-packet code, memory infidelity is suppressed to $O(γ^{r+1})$ while residual memory-probe correlations scale as $O(γ^{r+2})$. Finally, superconducting quantum circuit benchmarks under competing photon-number dephasing and Kerr nonlinearity identify preferred encoding regimes for hardware implementation.

quant-ph↗

arXiv2026

IMRouter: Imbalance-aware Temporal Pattern Routing for LLM-based Time Series Forecasting

Accurate time series forecasting (TSF) is essential for Web applications, such as traffic monitoring and online resource management. Recent advances in leveraging pre-trained large language models (LLMs) have shown promising performance in TSF. However, existing LLM-based TSF methods largely optimize the overall forecasting objective while overlooking temporal pattern imbalance (TPI), where different temporal patterns occur at highly unequal frequencies. Our empirical analysis reveals that TPI is prevalent across widely used TSF benchmarks, with substantially higher forecasting errors on less frequent patterns. To address this issue, we propose IMRouter (Imbalance-aware Temporal Pattern Routing), a novel framework that mitigates TPI-induced biases in both temporal representation learning and cross-modal alignment. Specifically, IMRouter introduces Pattern-Adaptive Temporal Representation Learning (PATR), which leverages gradient conflict and local gradient density to guide expert specialization. By penalizing incompatible expert assignments with density-aware weighting, PATR reduces the dominance of frequent optimization signals in router training. Furthermore, Routing-Guided Semantic Alignment (RGSA) transfers the pattern-specific knowledge learned by the router to cross-modal alignment through mutual information maximization, preserving distinctive temporal information across different patterns. Extensive experiments on benchmark datasets demonstrate the effectiveness of IMRouter, particularly in improving forecasting performance on less frequent temporal patterns.

cs.LG↗

arXiv2026

Glauber Description of Reaction Cross Sections with Bogoliubov Many-Body Wave Functions

I developed a microscopic framework that connects relativistic nuclear density functional theory to high-energy reaction observables by incorporating many-body Bogoliubov wavefunctions, represented in Pfaffian form, into the Monte Carlo integration of the Glauber model. This approach enables a simultaneous treatment of nuclear pairing correlations and multiple-scattering effects beyond the optical limit approximation (OLA). The method is applied to reaction cross sections of even-even O, Ne, and Mg isotopes on a $^{12}$C target. Complete multiple-scattering calculations systematically reduce the cross sections relative to the OLA and improves the agreement with measured interaction cross sections. Applications to the deformed Ne and Mg isotopes show that calculations based on a single intrinsic orientation depend strongly on the angle between the nuclear symmetry axis and the beam direction. Angular-momentum projection removes the spurious dependence on the arbitrarily selected intrinsic orientation and produces smooth isotopic systematics for the laboratory-frame cross sections. This work establishes a microscopic bridge between relativistic nuclear structure theory and reaction dynamics, providing a general framework for exploring exotic phenomena such as halos and clustering.

nucl-th↗

arXiv2026

FGD2HF: Frequency-guided Dual-domain Hypergraph Fusion for Semantic Segmentation

Frequency-domain information is beneficial for semantic segmentation. However, existing frequency-based methods often suffer from insufficient frequency utilization, inefficient fusion, and high computational complexity. To address these issues, we propose FGD2HF, a dual-domain fusion network that leverages frequency features to guide spatial-domain fusion and enhancement. Specifically, we introduce a novel Hybrid-Attention Encoder (NAHA) that jointly models local and global relationships through neighborhood and hypergraph attention, enhancing the network's ability to extract effective features. We further introduce a Wavelet Transform and Enhancement (WTE) module to extract and refine low- and high-frequency cues from preceding-stage features, followed by Frequency-guided Dual-domain Fusion (FGDF) modules that use frequency cues to condition current-stage spatial grouping and feature propagation. This design incorporates both global and local frequency guidance into spatial feature modulation, enhancement and fusion with low computational cost. Extensive comparisons with numerous state-of-the-art semantic segmentation methods demonstrate that our method is highly competitive. Our full FGD2HF-B model achieves 48.7% mIoU on ADE20K (+0.3% over SegMAN-B with 18% fewer GFLOPs), 83.2% mIoU on Cityscapes (comparable to SegMAN-B with 21% fewer GFLOPs), and 1.1% higher mIoU than VWFormer-T on SBD with 21% fewer GFLOPs, while also achieving significant improvements in boundary segmentation accuracy.

cs.CV↗

arXiv2026

EgoRayFormer: Ray-Space Stereo Refinement for Egocentric 3D Human Pose Estimation

Head-mounted stereo fisheye cameras provide a compact view of the wearer's body, but strong distortion, partial visibility and a short baseline make 3D pose estimation difficult. Existing refinement decoders sample image features independently in each view, without requiring the samples to correspond to a common 3D hypothesis. We introduce EgoRayFormer, a causal pose estimator that couples stereo feature sampling and geometric refinement in ray space. For each joint, the decoder projects shared 3D hypotheses along and around its viewing ray into both images. A geometric branch then corrects the projected joint locations and intersects the resulting rays; a learned gate blends this geometric estimate with a learned pose update. This formulation introduces an explicit stereo constraint while retaining flexibility when the rays provide weak depth constraints. In local evaluations on the official device-relative test splits, EgoRayFormer achieves three-seed mean MPJPEs of 17.64 mm on UnrealEgo2 and 49.55 mm on UnrealEgo-RW, improving to 47.57 mm with UnrealEgo2 pre-training. In paired validation ablations, removing the geometric branch increases MPJPE in all three seeds at both input resolutions. With a common ResNet-18 encoder and input resolution, our system reduces RW test MPJPE by 9.09 mm relative to the device-relative EgoPoseFormer baseline. A separate protocol analysis quantifies the difference between device-relative and oracle-root evaluation.

cs.RO↗

arXiv2026

LoongSage: A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes

Reinforcement learning (RL) has emerged as a critical phase in the post-training pipeline of large language models (LLMs). As the frontier of LLM capabilities shifts from static reasoning to agentic tasks--characterized by external tool invocation, multi-turn interactions, and long-horizon decision-making--existing mainstream RL frameworks have grown increasingly unwieldy, raising the overhead of adoption, usage, and debugging. To address these challenges, we present LoongSage, an agent-native, fully asynchronous RL framework with a lightweight core of roughly 14K lines of code. Built upon Megatron for distributed training and SGLang for high-throughput inference, LoongSage centralizes orchestration within a single controller process and encapsulates all GPU-resident components as Ray actors. It introduces AgentFlow, a unified execution paradigm that orchestrates both white-box and black-box agents. For asynchronous execution, it integrates sliding-window strategies that bound trajectory staleness, and employs TransferMesh, a topology-aware synchronization mechanism that routes weights after each step over the fastest transport available to each receiver. Beyond agentic RL, LoongSage incorporates a memory-efficient Teacher Manager that decouples the scaling of teacher models from GPU memory constraints, natively supporting Multi-Teacher On-Policy Distillation. Extensive evaluations across four complete training runs, spanning complex reinforcement learning and multi-teacher distillation, demonstrate that task rewards improve steadily while the train-inference gap is kept under continuous control, and the fully asynchronous pipeline removes synchronous idle bubbles, sustaining hardware utilization.

cs.LG↗

arXiv2026

Tempered fractional Laplacian: Boundary regularity

We study the boundary regularity of solutions for the tempered fractional Laplacian \[ (-Δ)^s_λu=f \quad \text{in }Ω, \qquad u=0 \quad \text{in }\mathbb{R}^n\setminusΩ, \] where $0 0$, and $Ω\subset\mathbb{R}^n$ is a bounded Lipschitz domain satisfying the exterior ball condition. The tempered operator is defined by \[ (-Δ)^s_λu(x) = c_{n,s}\operatorname{P.V.} \int_{\mathbb{R}^n} \frac{u(x)-u(y)}{|x-y|^{n+2s}} e^{-λ|x-y|}\,dy, \] which was originally proposed in [Koponen, Phys. Rev. E (1995), 1197-1199]. % Exponential tempering preserves the leading singularity of % the kernel but destroys its homogeneity, preventing a direct % application of classical boundary regularity estimates. % Extending these estimates [Ros-Oton and Serra, J. Math. Pures Appl. (2014), 275--302] to the tempered setting therefore % requires overcoming the difficulties caused by the loss % of scale invariance. By decomposing the operator into the classical fractional Laplacian and the correction operator $\mathcal Q_λ^s$, and combining approximation, localization, and iterative arguments with interpolation inequalities, we establish the global estimate \[ \|u\|_{C^s(\mathbb{R}^n)} \le C\|f\|_{L^\infty(Ω)}. \] For more regular data, we derive interior and boundary-weighted Schauder estimates. In particular, under suitable regularity assumptions on $f$, these estimates yield \[ |D^νu(x)| \le C\,δ(x)^{s-|ν|}, \quad |ν| > 0, \] where $ν$ is a multi-index and $δ(x)=\operatorname{dist}(x,\partialΩ)$. These bounds quantify the possible growth of higher-order derivatives near the boundary and provide a regularity foundation for numerical error analysis of the tempered problem.

math.FA↗

arXiv2026

Inverse identification of thermal-radiative parameters for reduced effective-temperature uncertainty in an ion optical clock

The blackbody-radiation (BBR) shift remains a leading systematic uncertainty of room-temperature single-ion optical clocks. Accurate evaluation of this shift requires the effective temperature at the ion position, which is difficult because RF heating, low and spatially varying surface emissivity and inaccessible trap surfaces make this quantity difficult to measure directly. Here, we developed a hybrid inversion framework that combines radiance-domain infrared thermography, black-painted high-emissivity reference regions, a finite-element model of a 171Yb+ end-cap ion trap and a reduced-order surrogate for repeated forward calculations and parameter identification. The surrogate reproduced the FE test fields with a full-field mean absolute error (MAE) of 0.02220 mK. The calibrated model predicted the ion-position effective temperature with a combined standard uncertainty of 155.0 mK, corresponding to a temperature-related BBR frequency-shift uncertainty of 1.647 x 10^-19 for the 2S1/2-2F7/2 (467 nm) E3 transition. This method provides a reproducible route for reducing and tracing the BBR uncertainty of room-temperature trapped-ion optical clocks.

physics.atom-ph↗

arXiv2026

Evo-RL: Iterative Robot Reinforcement Learning from Selective Experience

Existing real-world reinforcement learning methods improve robot policies by combining human corrections with autonomous rollouts. However, accumulated experience varies in its value for policy improvement, and even successful trajectories can contain inefficient or redundant behavior, motivating more selective experience reuse. To this end, we propose Evo-RL, an iterative reinforcement learning method that combines human interventions with advantage-guided experience selection. Evo-RL uses human corrections and experience with positive estimated advantages from successful trajectories to iteratively refine the policy. To support learning from these different experience sources, we introduce RW-RL-Dataset, which spans over 1,000 hours across four robot embodiments. On a real-world high-precision manipulation task, Evo-RL improves attempt success rate over RECAP by 30.46\%. Our code, model checkpoints, and dataset are publicly available.

cs.RO↗

arXiv2026

Precise description of spontaneous fission half-lives of Rutherfordium isotopes within the relativistic Hartree-Bogoliubov theory

I investigate the spontaneous-fission (SF) half-lives of $^{254,256}$Rf within the multidimensionally constrained relativistic Hartree-Bogoliubov (MDC-RHB) model combined with the local quasiparticle random-phase approximation (LQRPA) method. The collective inertia is evaluated using the finite-amplitude method (FAM), which enables a fully self-consistent treatment of dynamical residual effects beyond the conventional cranking approximation. The effective collective potential is constructed by subtracting both the vibrational zero-point energy and rotational energy correction from the deformation energy surface. The SF half-lives are evaluated within the Wentzel-Kramers-Brillouin (WKB) approximation. Compared with the nonperturbative and perturbative cranking approximation, the FAM-LQRPA collective inertia is significantly enhanced, especially around the ground-state minimum, due to the dynamical residual interaction included in the QRPA. As a consequence, the predicted SF half-lives increase by several orders of magnitude and become consistent with experimental data. The calculated SF half-lives are 5.65 $μ$s for $^{254}$Rf and 0.95 ms for $^{256}$Rf, both within approximately two orders of magnitude of the corresponding experimental values. These results indicate that a self-consistent treatment of dynamical residual effects and beyond mean-field correlations is crucial for quantitative microscopic descriptions of SF within covariant density functional theory.

nucl-th↗

arXiv2026

Generation of large-amplitude squeezed cat states with near-unity efficiency

The Gottesman-Kitaev-Preskill encoding has emerged as a leading candidate for fault-tolerant quantum computation with continuous variables. For photonic architectures, the major challenge is preparing high-quality resource states, which can be deterministically synthesised from many large-amplitude cat states. Thus far, only modest-sized optical cat states have been prepared experimentally, and the implemented methods are highly probabilistic. We propose an all-optical scheme utilising quantum non-demolition interactions and photon-number measurements to prepare large-amplitude cat states with near-unity probability. Importantly, no particular photon-number outcome is postselected: all outcomes contribute to the accumulated photon number, with the protocol repeated until a required threshold is reached. We demonstrate that the scheme is robust to realistic levels of photon-loss, the dominant source of error in optical systems. Our results highlight the power of active Gaussian operations for state preparation and pave the way for efficient quantum error correction using bosonic codes.

quant-ph↗

arXiv2026

Quantum advantage in learning single mode bosonic channels

Quantum resources can dramatically reduce the data required to learn physical systems, with exponential improvements in sample complexity demonstrated in several quantum learning tasks. However, these advantages have typically relied on quantum resources that scale with problem complexity, most notably increasing system dimension or entanglement. This raises a fundamental question: can exponential quantum learning advantages arise within a fixed, unentangled quantum system? In this paper, we address this question by considering the learning of an unknown random-displacement distribution whose complexity is determined not by the dimensionality of the physical system, but by the Fourier resolution of its features. We show that the quantum-limited noise of vacuum probes progressively obscures high-frequency features, leading to an exponential growth in sample complexity. We establish an information-theoretic lower bound for arbitrary classical-state probes and show that squeezing overcomes this classical limit by extending the accessible Fourier bandwidth. Experimentally, we demonstrate an exponential reduction in sample complexity using squeezed vacuum probes for both binary hypothesis testing and characteristic-function reconstruction. Our results show that a single bosonic mode can exhibit exponential quantum learning advantages without entanglement or an increase in system size, identifying accessible Fourier bandwidth as a fundamentally distinct resource for quantum-enhanced learning.

quant-ph↗

arXiv2026

Investigation of Nonlinear Collective Dynamics in Relativistic Heavy-Ion Collisions Using A Multi-Phase Transport Model

The nonlinear response coefficient, $χ_{4,22}$, is a crucial observable for probing the dynamical properties of the quark-gluon plasma (QGP). Although traditionally interpreted as a signature of medium response, recent studies suggest that $χ_{4,22}$ also encodes information about the intrinsic initial-state configuration of the colliding nuclei. In this study, we use the A Multi-Phase Transport (AMPT) model to investigate the microscopic origin and stage-by-stage development of $χ_{4,22}$ in $^{238}$U+$^{238}$U and $^{197}$Au+$^{197}$Au collisions at $\sqrt{s_{\rm NN}} = 200$ GeV. By tracking flow observables through the partonic cascade, quark coalescence, and hadronic rescattering phases, we map the conversion of initial geometric eccentricities into final-state momentum anisotropies. Our results demonstrate that the absolute magnitude of $χ_{4,22}$ increases continuously during collective expansion, confirming its nature as a dynamically generated medium response. In contrast, the relative ratio of this coefficient between the U+U and Au+Au systems, $R(χ_{4,22})$, exhibits approximate stage independence over a broad centrality range, as quantified by a constant-fit test across the three AMPT evolution stages. This indicates that the ratio reduces common medium-response effects, such as overall amplification efficiency and viscous attenuation, and therefore retains stronger sensitivity to the relative initial-state geometry. These findings provide theoretical support for using nonlinear flow ratios to constrain higher-order nuclear structure, such as hexadecapole deformation, while also clarifying the residual stage dependence and nonflow limitations relevant to precision extractions.

nucl-th↗

arXiv2026

MV-STRIDE: Enabling MLLMs to Master Multi-View Spatial Reasoning via Hierarchical Capability Modeling

Despite the rapid progress of Multimodal Large Language Models (MLLMs) in 2D vision-language tasks, robust multi-view spatial reasoning remains a fundamental bottleneck due to the lack of structured 3D cognitive pathways in existing datasets. To address this, we introduce MV-STRIDE, a Multi-View hierarchical SpaTial Reasoning dataset with Interdependent and DEcomposed capabilitiEs. Moving beyond flat data structures, MV-STRIDE explicitly models the dependency relationships between foundational perception, scene understanding, and complex contextual reasoning, providing a coherent learning pathway aligned with human spatial cognition. We develop a systematic QA generation pipeline leveraging diverse 3D scene sources that enforces cross-view dependency constraints to prevent single-view solvability, generating multi-level spatial reasoning tasks supported by cognitively grounded chain-of-thought supervision for complex inference. Extensive evaluations demonstrate that our multi-stage training framework based on our hierarchical dataset achieves state-of-the-art performance across multiple spatial reasoning benchmarks, notably the multi-view oriented MMSI-Bench. Our approach enables MLLMs to maintain robust, 3D-consistent spatial reasoning across diverse viewpoints. The code and dataset are available at https://co1dspring.github.io/MV-STRIDE/.

cs.CV↗

arXiv2026

Modeling and Resource Optimization for Quantum Oracles

Quantum oracles are fundamental building blocks of many quantum algorithms, and their resource consumption directly affects performance, yet structured description and complexity analysis for their composition are still lacking. In this paper, we introduce the Framework for Oracle Recursion Modeling (FORM), a unified formal abstraction of multi-function composition in quantum oracles: it provides a structured description of the composition layer, makes its gate complexity exactly computable, and turns oracle design into an optimizable tree-construction problem. Based on this model, we propose the ShallowGrow algorithm, which constructs an oracle structure under a given ancilla budget and provably minimizes the number of function evaluations. On Boolean quadratic equation systems, ShallowGrow reduces Qiskit-measured circuit depth by 54.1% on average relative to the state-of-the-art W-cycle construction, with consistent reductions on the EPFL and ISCAS85 combinational logic networks under scarce ancilla budgets. Furthermore, pebbling-based syntheses trade space against time within the logic network of a function; ShallowGrow extends this trade-off across functions, and integrated with their published circuits it reduces the ancillary qubits of a complete oracle from one per constraint function to logarithmically many. With half as many ancillas as constraint functions, the T-count falls by a factor of 7.9 to 32 relative to the W-cycle-based construction.

quant-ph↗