arXiv subjects

Yuan Lu

Publications and source records attributed to Yuan Lu.

At least 19 recordsLinked to original sources

arXiv2026

Harness-S: Learning from Sibling Harnesses for Generalizable Agent Training

Harness evolution from a common base exhibits Locality, Complementarity, and Nonstationarity: variants repair different failures, and their utility changes as the policy learns. Rather than seeking a universally superior harness, we introduce Harness-S to selectively rescue failures of a base execution system through sibling harnesses and transfer the resulting capabilities into a single policy that remains effective under the original harness. Directional Harness Evolution develops specialized sibling harnesses along distinct repair directions, forming a reusable Sibling Harness Pool that remains fixed during policy training. Complementary Harness Rescue Training invokes this pool only when all initial attempts under the base harness fail. The same policy generates fresh rollouts under selected auxiliary harnesses, and these trajectories replace the failed group for reinforcement learning. An Adaptive Harness Router selects the Top-k variants using Harness Behavior Profiles and voting across multiple LLMs. Policy-aware profile re-anchoring refreshes this evidence as the policy learns. Across AppWorld, SWE-Gym, and DBBench, the trained policy substantially outperforms policy-only training when returned to the original base harness. Behavioral analyses and held-out cases support the transfer of harness-elicited behaviors into the policy itself. Reusing the router for Rank-1 inference provides additional gains at the same executor rollout count.

cs.AI↗

arXiv2026

Foundation Models to Unlock Real-World Evidence from Nationwide Medical Claims

Real-world data increasingly inform regulatory and clinical decisions. Administrative claims, among their major sources, continuously track care and its cost across providers throughout enrollment, yet remain largely unexplored for healthcare foundation models. Here we present ReClaim, a family of generative foundation models built on 43.8 billion claims records from more than 200 million enrollees in the MarketScan Research Databases. ReClaim learns diagnoses, procedures, medications and expenditure jointly through pre-training from scratch, then specializes for downstream prediction tasks through post-training, with a training cutoff at the end of 2022. In predicting more than 1,000 diseases, ReClaim achieved a mean ROC-AUC of 0.769, outperforming disease-specific gradient-boosted models (0.663) and a transformer-based baseline (0.694), and retained this advantage when generalizing to two independent cross-domain electronic health record datasets. ReClaim also outperformed gradient-boosted models in forecasting next-year expenditure and in identifying high-need, high-cost enrollees. ReClaim maintained its advantage in disease prediction and expenditure forecasting on post-cutoff data (2023-2024). Beyond prediction, ReClaim representations used as covariates resulted in improved effect estimates compared with algorithm-derived covariates, when benchmarked against results expected from randomized controlled trials. Together, these results establish nationwide insurance claims as a scalable substrate for healthcare foundation models supporting disease prediction, expenditure forecasting and confounding adjustment in real-world evidence studies. Code is available at https://github.com/BIDS-Xu-Lab/ReClaim.

cs.AI↗

arXiv2026

Probing fluctuations in photon emission statistics of spin injected quantum dots with time dependent electronic excitation

Electrically spin-injected ensembles of InGaAs/GaAs quantum dots emit circularly polarized single photons, with the emission helicity determined by the spin polarization of the injected electrons. A ferromagnetic CoFeB spin-injection layer with perpendicular magnetic anisotropy, combined with an MgO tunnel barrier, enables spin-polarized electron injection into the quantum-dot active region, resulting in circularly polarized photon emission at zero applied magnetic field. These devices realize direct spin-charge-photon conversion in an electrically driven, microelectronics-compatible quantum light source that operates without an external pump laser. Using an extended Hanbury Brown-Twiss interferometer with four detectors and time-tagged readout, we measure time-resolved photon-number statistics that resolve up to $|2\rangle$ Fock states within a single excitation cycle. While single-photon ($n=|1\rangle$) emission tracks the electrical excitation pulse as expected, the two-photon ($n=|2\rangle$) statistics exhibit pronounced fluctuations at multiple temporal positions within the pulse. These fluctuations vanish upon demagnetization of the spin-injection layer, establishing a direct link between the multi-photon emission dynamics and the spin polarization of the injected carriers. The observations indicate that spin injection controls not only the helicity of the emitted photons but also higher-order temporal correlations of the emission. The behavior is phenomenologically captured by inserting both a variable excitation mechanism and a set of multi-emitter collapse operators within a stochastic wave function evolution. The platform thus opens a route to electrically driven, spin-controlled single and multi-photon Fock-state sources and provides a testbed for spin-conditioned multi-emitter dynamics in solid-state quantum light sources.

quant-ph↗

arXiv2026

CleanRM: Reward Modeling from Noisy Feedback with a Few Clean Feedback for RLHF

Feedback noise is pervasive in preference data, arising from subjective human judgments, annotation mistakes, and imperfect LLM judges. Standard reward modeling trained on such feedback yields a biased estimate of the clean risk. Although classical denoising methods are directly applicable to noisy feedback learning, they generally assume a uniform noise ratio across samples. However, this assumption fails for feedback data, where noise is intrinsically sample-wise. To address this challenge, we propose CleanRM, which uses a few clean feedback to calibrate sample-wise noise ratios via optimal transport. Specifically, CleanRM aligns the joint distribution of preferences reverted from noisy feedback with that of clean feedback and then trains the reward model using a noise-corrected surrogate loss. Theoretically, we prove that the resulting objective is unbiased with respect to the clean risk. Extensive experiments demonstrate that CleanRM outperforms state-of-the-art denoising baselines and yields more effective downstream RLHF alignment.

cs.AI↗

arXiv2026

Rubric Judgments Lack Uncertainty: Conformal Weighting for LLM Evaluation

Rubric-based evaluation is widely used to assess large language model (LLM) outputs by decomposing quality into multiple criteria and assigning each a binary judgment from an LLM verifier. However, judgment certainty varies substantially across criteria. Existing methods uniformly aggregate these judgments into a final score, implicitly treating every criterion as equally certain and overlooking this heterogeneity. As a result, uncertain judgments remain incorporated into reward estimation without distinction, potentially yielding misleading training signals for downstream reinforcement learning (RL). To address this limitation, we propose Uncertainty-aware Rubric Weighting (URW), a training-free framework that quantifies criterion-level judgment uncertainty via conformal calibration. Specifically, URW elicits a soft uncertainty signal from a fixed pool of paraphrased judgment templates and converts it into a per-criterion weight using split-conformal $p$-values with provable coverage guarantees. Extensive experiments demonstrate that URW consistently improves pairwise preference accuracy and delivers stronger downstream RL alignment.

cs.AI↗

arXiv2026

Deep Time-series Forecasting Needs Kernelized Moment Balancing

Deep time-series forecasting can be formulated as a distribution balancing problem aimed at aligning the distribution of the forecasts and ground truths. According to Imbens' criterion, balancing two distributions requires that their first moments be identical with respect to any balancing function. We demonstrate that existing objectives enforce moment matching only for one or two predefined balancing functions, thus failing to achieve full distribution balance. To address this limitation, we propose direct forecasting with kernelized moment balancing (KMB-DF). Unlike existing objectives, KMB-DF adaptively selects the most informative balancing functions from a reproducing kernel Hilbert space (RKHS) to enforce sufficient distribution balancing. A tractable and differentiable formulation is derived to enable estimation from empirical samples and seamless integration into gradient-based training pipelines. Extensive experiments across multiple models and datasets show that KMB-DF consistently improves forecasting accuracy and achieves state-of-the-art performance. Code is available at https://anonymous.4open.science/r/KMB-DF-403C.

cs.LG↗

arXiv2026

Not All Criteria Matter: Importance-Aware Criteria Weighting for Rubric-Based LLM Evaluation

Rubric-based evaluation assesses the responses of large language models (LLMs) by aggregating the verification results of explicit criteria, yet not all criteria matter equally. However, existing methods commonly aggregate the verification results with uniform weighting, which assigns minor and overlapping criteria the same weight as essential ones and thus produces misleading evaluation scores. Although some methods prompt an LLM to directly assign weights, direct weighting requires a holistic judgment over all criteria that LLMs struggle to make reliably. To address these limitations, we propose Importance-Aware Criteria Weighting (ICW), which derives the weights of criteria from pairwise importance comparisons made by an LLM. Specifically, ICW transforms the holistic judgment over all criteria into a series of pairwise importance comparisons, checks and revises inconsistent comparisons, and derives the weights from the principal eigenvector of the comparison matrix. Extensive experiments demonstrate that ICW consistently outperforms uniform weighting and competitive weighting methods, and translates into better policy models in downstream reinforcement learning.

cs.CL↗

arXiv2026

FleetPilot: Fine-Grained Steering Empowers Real-World LLM Personalization

With the rapid evolution of large language models (LLMs), there is an increasing demand for effective personalization techniques that can adapt model behavior to individual user preferences. Personalized LLMs must infer a user's query-dependent preferences from limited, heterogeneous, and evolving interaction histories. Existing approaches rely primarily on in-context personalization, parametric adaptation, or personalized reward modeling, but respectively incur recurring costs in context management, model optimization and deployment, or multi-candidate generation and scoring. These limitations become particularly restrictive when user observations are sparse and the preference evidence relevant to each query varies across contexts and over time. To address these challenges, we propose \modelname, a fine-grained and instance-tailored steering framework that dynamically generates sample-level interference vectors from user data and injects them into the model's forward pass for personalized adaptation. Our approach introduces: match contrast that produces user-relevant signals, fine-grained steering component that captures nuanced signals by hooking activations from attention and MLP layers, and query-conditioned aggregation module that synthesizes these signals into contextually relevant enhancements. Extensive experiments across diverse scenarios, including short-to-long text generation, and web function calling, validate the effectiveness and compatibility of \modelname. Results show that our method consistently brings significant performance gains, and show extreme superiority in fast-shifting and data-sparse environments. Additionally, \modelname rids us of training cycles and incurs light overhead of storage or inference latency. These make \modelname a strong and real-world deployable method. Implementation is available \footnote{https://anonymous.4open.science/r/FleetPilot}.

cs.AI↗

arXiv2026

Symmetry-Dependent Polarity Reversal of Bulk Spin-Orbit Torque in Single-Layer MnCoGa

Controlling the magnitude and polarity of spin-orbit torque (SOT) is essential for manipulating magnetization in spintronic devices. Here, we realize bulk spin-orbit torque (SOT) in a single-layer Mn1.6Co1.4Ga (MCG) Heusler alloy and achieve lattice strain-controlled reversal of SOT polarity. SOT-driven magnetization switching is achieved over a thickness range of 3-20 nm, with the switching polarity reversed between the strongly strained ultrathin regime and thicker films. First-principles calculations reveal a sizable intrinsic spin Hall conductivity (SHC) in MCG originating from its topological band structure, which is enhanced by the tetragonal distortion without changing sign. Together with the thickness-dependent structural evolution, we demonstrate that the reversal of SOT polarity does not originate from a sign change of the SHC, but rather from lattice distortion induced symmetry breaking that modifies the conversion of spin current into a net bulk SOT, thereby controlling its polarity. These results establish strain-controlled structural symmetry as an additional degree of freedom for tuning bulk SOT in magnetic single layers.

cond-mat.mes-hall↗

arXiv2026

Self-induced spin-orbit torque switching in a synthetic antiferromagnetic Co 2 MnGa /MnGa bilayer

The large intrinsic spin current in magnetic Weyl semimetals (WSMs) provides a promising platform for spin-orbit torque (SOT) devices. Here, we demonstrate SOTdriven magnetization switching in a synthetic antiferromagnet (SAF) composed of a Co2MnGa(CMG)/MnGa bilayer. In this heavy metal-free structure, CMG film functions simultaneously as a spin current source and as a magnetic layer. Macrospin simulations confirm the self-induced switching mechanism in the bilayer. The SOT generated by CMG switches the MnGa layer, which in turn triggers the reversal of the CMG layer via strong antiferromagnetic exchange coupling. These results elucidate the spin dynamics in SAFs and provide a new pathway toward SOT devices with selfinduced magnetization switching.

cond-mat.mes-hall↗

arXiv2026

Programmable Asymmetric Spin-Orbit Torque Switching for Spin Logic

Spin logic devices provide a promising route toward ultralow-power and nonvolatile information processing. In this work, we demonstrate field-free spin-orbit torque-induced asymmetric magnetization switching in in-plane anisotropy (IMA) and perpendicular magnetic anisotropy heterostructures, characterized by different critical switching currents under opposite current polarities. Combined experiments and macrospin simulations reveal that the asymmetric switching originates from an in-plane effective field Hy , which breaks the Myz mirror symmetry of the system. By controlling the magnetization direction of the IMA layer, both the switching polarity and bias direction can be tuned. Building on four distinct types of asymmetric switching behaviors, we realize a reconfigurable spin logic operation within a single Hall-bar device. These results clarify the physical origin of asymmetric magnetization switching and demonstrate an approach for realizing reconfigurable spin logic devices.

cond-mat.mes-hall↗

arXiv2026

DistDF: Time-Series Forecasting Needs Joint-Distribution Wasserstein Alignment

Training time-series forecasting models requires aligning the conditional distribution of model forecasts with that of the label sequence. The standard direct forecast (DF) approach resorts to minimizing the conditional negative log-likelihood, typically estimated by the mean squared error. However, this estimation proves biased when the label sequence exhibits autocorrelation. In this paper, we propose DistDF, which achieves alignment by minimizing a distributional discrepancy between the conditional distributions of forecast and label sequences. Since such conditional discrepancies are difficult to estimate from finite time-series observations, we introduce a joint-distribution Wasserstein discrepancy for time-series forecasting, which provably upper bounds the conditional discrepancy of interest. The proposed discrepancy is tractable, differentiable, and readily compatible with gradient-based optimization. Extensive experiments show that DistDF improves diverse forecasting models and achieves leading performance. Code is available at https://anonymous.4open.science/r/DistDF-F66B.

cs.LG↗

arXiv2026

Quadratic Direct Forecast for Training Multi-Step Time-Series Forecast Models

The design of learning objectives is central to training time-series forecasting models. Existing learning objectives such as mean squared error mostly treat each future step as an independent, equally weighted task, which leads to the following two challenges: (1) they overlook the label autocorrelation effect among future steps, leading to biased learning objectives; (2) they fail to set heterogeneous task weights for different forecasting tasks corresponding to varying future steps, limiting the forecasting performance. To fill this gap, we propose a novel quadratic-form weighted learning objective, addressing both issues simultaneously. Specifically, the off-diagonal elements of the weighting matrix account for the label autocorrelation effect, whereas the non-uniform diagonals are expected to match the preferred weights of the forecasting tasks with varying future steps. On this basis, we propose a Quadratic Direct Forecast (QDF) learning algorithm, which trains the forecast model using the adaptively updated quadratic-form weighting matrix. Experiments show that our QDF effectively improves the performance of various forecast models, achieving state-of-the-art results. Code is available at https://github.com/Master-PLC/QDF.

cs.LG↗

arXiv2026

Deep Time-Series Forecasting in 10 Years: A Survey

Autocorrelation is a common property of time-series, where each observation is dependent on its predecessors. In deep time-series forecasting, it raises two central challenges: (1) designing backbone architectures to model autocorrelation in history sequences, and (2) devising loss functions to model autocorrelation in label sequences. Recent studies have made strides in tackling these challenges, but a systematic survey examining both aspects remains lacking. To bridge this gap, this paper reviews deep time-series forecasting from an autocorrelation modeling perspective, offering two contributions beyond existing surveys. First, it introduces a taxonomy that jointly covers both backbone architectures and loss functions, whereas prior surveys provide limited coverage of the latter. Second, it analyzes the motivations and insights underlying the surveyed literature from a unified autocorrelation perspective, providing a holistic overview of the field's evolution. Additional resources and details are available at https://github.com/Master-PLC/Awesome-TSF-Papers.

stat.ML↗

arXiv2026

Unbiased Reward Modeling from Implicit Feedback for LLM Alignment

Despite the success of reinforcement learning from human feedback (RLHF), existing reward modeling methods largely rely on explicit feedback, which is costly to collect and difficult to scale. This work studies implicit reward modeling, learning reward models from implicit user feedback, such as clicks, copies and skips. While scalable and cost-effective, implicit feedback poses two key challenges: It lacks definitive negative samples, which makes standard positive-negative classification methods inapplicable; It suffers from selection bias, where responses have heterogeneous propensities to elicit feedback, which further obscures definitive negative samples. To address these challenges, we propose ImplicitRM, which learns unbiased reward models from implicit feedback. It stratifies training samples into four latent groups using a stratification model and derives a likelihood-maximization objective that is theoretically unbiased, thereby addressing both challenges. Experiments across diverse LLM backbones and benchmark datasets validate that ImplicitRM learns accurate reward models from implicit feedback and improves performance on downstream RLHF tasks.

cs.CL↗

arXiv2026

Rubrics on Trial: Evolving Rubrics from a Single Query via Synthetic Pairwise Evidence

Rubric evolution offers a promising approach to improving the quality of rubrics generated by large language models (LLMs). Central to this process is rubric comparison, which identifies the better of two rubrics and guides the direction of evolution. However, accurate rubric comparison is difficult, which presents two challenges. (1) It should reflect downstream task performance, which is essential for assessing rubric utility but often prohibitively expensive to evaluate. (2) It should discourage unnecessary criteria, which increase verification costs and may dilute the influence of essential criteria. To address these challenges, we introduce Rubrics on Trial, a multi-agent framework that evolves rubrics by comparing synthetic response pairs. To address challenge 1, the framework compares synthetic responses that satisfy the respective rubrics, providing a proxy for downstream performance without training a separate policy for each rubric. To address challenge 2, it assesses the necessity of a candidate criterion by independently generating high-quality alternative responses that violate it and comparing them with edited versions that satisfy it. A rubric is favored when it improves response quality in both comparisons, and the resulting comparison signal is further incorporated for rubric evolution. Extensive experiments demonstrate that Rubrics on Trial improves the quality of generated rubrics and leads to better downstream task performance.

cs.CL↗

arXiv2026

RubricArmor: Adversarial Evolution Improves LLM-Based Rubric Generation

Rubric-based reinforcement learning (RL) provides interpretable rewards for aligning large language models (LLMs) by evaluating responses against query-specific evaluation criteria. To construct rubrics at scale, a straightforward approach to LLM-based rubric generation is to prompt an LLM to generate a rubric directly from the query. However, rubrics directly generated by LLMs are vulnerable to reward hacking, since omitted or underspecified criteria allow the policy to obtain high rubric rewards with low-quality responses. Existing LLM-based rubric generation methods improve the granularity and coverage of the generated criteria but do not proactively guard against reward hacking. To address this limitation, we propose RubricArmor, an adversarial framework that exposes and mitigates potential reward hacking at the rubric generation stage before it occurs in subsequent RL. Specifically, RubricArmor performs adversarial evolution, in which an attack step and a repair step alternate over multiple rounds. The attack step simulates the reward hacking of the policy by constructing adversarial responses that satisfy the current rubric but fail to properly complete the task. The repair step then revises the rubric to detect the response defects exposed by the attack step while preserving other valid criteria. Extensive experiments demonstrate that RubricArmor outperforms competitive rubric generation baselines and translates into more effective downstream rubric-based RL.

cs.CL↗

arXiv2026

Learning Where to Steer: Noise-Space Geometry for Efficient Offline Multi-Objective Optimization with Generative Models

Offline multi-objective optimization (MOO) seeks solutions with better objective trade-offs using only a fixed dataset, without querying the objectives. Diffusion models trained on such data have emerged as a promising approach, but their samples are not inherently better than the data and must be steered toward the Pareto front. Existing methods guide or condition every sampling step. We instead act on the initial noise and leave the sampling process unchanged. Across Off-MOO-Bench, we observe that the objectives, as functions of the noise, are sensitive to only a few directions. We estimate these directions once per task via a Recursive Feature Machine using function values alone, and a small cache serves every trade-off, so each candidate costs one noise displacement and one ODE solve. We prove that this displacement increases the learned scalarized objective in expectation, and that sweeping trade-offs recovers the flow's attainable front up to proxy and steering errors. With additional guidance, for which we introduce novel data-adaptive and Pareto-aware operators, our method attains the best average hypervolume rank among generative methods on 47 tasks, at comparable or lower sampling cost. Steering alone outranks the best prior generative method at a fraction of its sampling cost.

cs.LG↗