Searcharxiv⌕ Search

arXiv subjects

Yan Xu

Publications and source records attributed to Yan Xu.

At least 37 records · Page 2Linked to original sources
arXiv2026

BALTO: Balanced Token-Level Policy Optimization for Hallucination Mitigation

Hallucinations remain a major obstacle to deploying large language models (LLMs) in knowledge-intensive settings, where generated responses must be faithfully grounded in provided evidence. Reinforcement learning (RL) is a promising direction for hallucination mitigation, but response-level faithfulness rewards suffer from a granularity mismatch: localized hallucinations can cause supported content to receive spurious penalties. Although recent work introduces fine-grained feedback such as claim-level verification and token-level rewards, unbalanced credit assignment can still induce length, verbosity, or optimization-noise biases. We propose BALTO, a Balanced Token-level Policy Optimization framework for hallucination mitigation. BALTO extracts checkable factual claims, verifies them against the reference context, and projects claim-level judgments to token-level labels. A balanced token-level credit assignment mechanism is introduced into the framework. This design redistributes probability mass from unsupported content toward faithful content, rather than suppressing the entire response. We systematically analyze the limitations of response-level rewards from a theoretical standpoint, and prove BALTO's advantages in training stability and optimization efficiency for hallucination mitigation. Experiments on ConFiQA, RAGTruth, and FinLLM-Eval show that BALTO achieves the highest faithfulness across all six model--benchmark settings and consistently outperforms existing post-training baselines in Q-Score, demonstrating a stronger faithfulness--informativeness trade-off.

cs.CL↗
arXiv2026

Prediction of Solar Flares Using Photospheric Magnetic Field Parameters with Deep Learning

Solar flares, particularly those of the M- and X-class, have a significant impact on human life because of their potential to disrupt critical infrastructure and communication systems on Earth. Accurate prediction of solar flares is crucial for mitigating these risks, but the black-box nature of conventional deep learning models used in flare prediction limits their trustworthiness and interpretability. In this paper, we propose a new approach to solar flare prediction using photospheric magnetic field parameters or features with deep learning. To improve model interpretability, we integrate explainable artificial intelligence (XAI) techniques, including SHapley Additive exPlanations (SHAP) and partial dependence plots (PDPs), into our prediction framework. XAI methods provide transparency by analyzing the importance and interactions of features used by our model. Specifically, SHAP values offer a global and local understanding of the features, while PDPs provide insights into feature-level trends. These techniques demonstrate the potential of XAI in deploying AI-driven solutions in high-impact applications such as solar flare prediction, paving the way for more informed decision-making in solar physics and space weather studies.

astro-ph.SR↗
arXiv2026

Decision-Focused Continual Learning for Seaport Power-Logistics Scheduling: Generalization across Varying Tasks

Power-logistics scheduling in modern seaports typically follows a predict-then-optimize pipeline. To enhance the decision quality of predictions, decision-focused learning has been proposed, which aligns the training of forecasting models with downstream decision outcomes. However, this end-to-end design inherently restricts the value of forecasting models to a specific task structure and therefore generalizes poorly to evolving tasks induced by varying vessel arrivals. We address this gap with a decision-focused continual learning framework that adapts online to a stream of scheduling tasks. Specifically, we introduce Fisher-information-based regularization to enhance cross-task generalization by preserving parameters critical to prior tasks. A differentiable convex surrogate is also developed to stabilize gradient backpropagation. The proposed approach enables learning a decision-aligned forecasting model across a varying task stream with sustainable long-term computational and memory requirements. Experiments calibrated to Jurong Port show improved decision performance and cross-task generalization over existing methods, together with reduced computational cost and a bounded memory footprint.

cs.LG↗
arXiv2026

UniPET: a universal network for high-quality PET image denoising across varied dose reduction factors

Most existing deep learning-based PET image denoising methods assume a fixed and known dose reduction factor (DRF) for low-dose PET images. However, these methods encounter significant performance degradation when the DRF varies beyond the assumed one in practical applications. To address the challenge posed by varied DRFs, several preliminary studies focus on the task of universal PET image denoising, aiming to train a universal model over low-dose data across DRFs. Nonetheless, these vanilla universal models often struggle with misaligned styles present in different DRF data, leading to the \textit{style elimination issue} with a significant over-smoothing effect. To deal with this issue, we innovatively introduce domain generalization to PET image denoising and propose a universal PET image denoising network (UniPET) to achieve high-quality PET image denoising across diverse DRFs. UniPET comprises two primary innovations: a style alignment network (SAN) and a region-aware learning strategy (RALS). Specifically, SAN utilizes style alignment techniques derived from domain generalization to align and recover styles across different DRFs, ensuring the model's generalizability across various DRFs while effectively preserving styles. Furthermore, to enhance style recovery, RALS distinguishes between flat and stylized regions, exclusively conducting adversarial learning on the latter, thereby more effectively guiding the model's focus towards learning stylized regions. It is demonstrated that our proposed UniPET can adaptively recover different DRF styles and achieve high-quality PET image denoising across DRFs. Comprehensive experiments show that UniPET exhibits comparable performance to individual DRF-specific models at specific DRFs and realizes state-of-the-art performance in universal PET image denoising quantitatively, perceptually, and clinically.

cs.CV↗
arXiv2026

Power Grid Topology Control

Power grids are facing major challenges from growing renewable integration and worsening climate impacts. While flexibility on both the demand and generation sides has been widely explored to address these challenges, network-side flexibility, especially in network topology, remains highly underutilized. Advances in communication, power electronics, and circuit breakers have made network topology increasingly controllable. However, leveraging this topological flexibility poses substantial challenges, primarily due to the inherent non-convexity and hybrid dynamics in associated optimization and control problems. This monograph surveys the development of power grid topology control in both early and recent years. It begins by discussing the fundamental topological constraints involved in topology control problems. Subsequently, it introduces steady-state topology control for transmission and distribution networks separately, covering fundamentals, a state-of-the-art review, and representative recent advances. Additionally, the network topology transition problem, which addresses the implementation of optimal topology solutions and has garnered increasing attention in recent years, is further modeled and analyzed. Beyond utilizing the flexibility of steady-state network topology, controlling network topology during transients can also contribute to system stabilization. Traditional approaches, such as intentional controlled islanding for transmission networks, as well as recently developed topology control methods for microgrid stabilization, exemplify this concept. Finally, a summary of this monograph is provided.

eess.SY↗
arXiv2026

Comprehensive Analysis of Optical brightness and Color Variability of Blazars in the ZTF Survey DR22

This study conducts a comprehensive analysis of brightness and color variability in blazars, utilizing over six years of quasi-simultaneous g-band and r-band data from 1149 sources in the ZTF Data Release 22 (DR22), including 589 BL Lacs and 560 FSRQs. We quantify the amplitude of variability and the fractional root mean square (rms) variability for each source and statistically assess the overall and short-term color behaviors across different subclasses; examine the distribution of brightness variability characteristics across different blazar types and investigate how the extent of variability correlates with color trends. We found BL Lacs tend to exhibit a BWB (bluer when brighter) trend, while FSRQs display a RWB (redder when brighter) trend; BL Lacs with negligible host-galaxy contamination exhibit a BWB trend fraction of 14.7% (68/462) compared to 2.3% (11/462) for RWB trend, while FSRQs show 8.8% (49/560) BWB trend versus 14.1% (79/560) RWB trend. By statistically investigating how color behavior depends on brightness state across different timescales, we find that brighter states in both BL Lacs and FSRQs are more likely to exhibit BWB trend. Our results also show that BL Lacs with a BWB trend exhibit higher variability than those with a RWB trend, whereas FSRQs with a RWB trend display significantly greater variability than those with a BWB trend. These results suggest that blazar color variability depends jointly on source type, brightness state, and variability amplitude, highlighting the complexity of color evolution in blazars.

astro-ph.CO↗
arXiv2026

Coupling Spectrum Estimation and Single Energy Material Decomposition via X-ray Grating Interferometry

Dark-field imaging based on grating interferometry is an emerging X-ray modality in medical imaging, which is particularly capable of providing complementary diagnostic information by visualizing the microstructural properties of lung tissue. However, quantitative dark-field imaging remains fundamentally challenged by beam hardening, which arises from the energy-dependent fringe visibility under polychromatic illumination. The resulting artifacts substantially degrade the quantitative accuracy of dark-field images. In this work, motivated by our key observation of an intrinsic similarity between the X-ray energy spectrum and the system-related coupling spectrum, we propose a unified framework to simultaneously and independently estimate both spectra. By measuring the transmission associated with the zeroth- and first-order components of the phase-stepping curve using solid step-wedge phantoms, the two spectra are robustly estimated via an expectation-maximization algorithm. The recovered spectra are subsequently incorporated into a physics-based correction model to mitigate beam-hardening-induced artifacts in dark-field imaging effectively. Furthermore, leveraging the inherent availability of two independent spectra within X-ray grating interferometry, we introduce a single-energy material decomposition method that achieves basis material imaging without dual-energy scans. Wave-optical simulations and experiments demonstrate accurate spectrum estimation, effective dark-field signal correction, and reliable material decomposition. Consequently, the proposed framework extends the diagnostic potential of X-ray grating interferometry beyond pulmonary imaging, facilitating broader applications in medical imaging.

physics.med-ph↗
arXiv2026

Anticipate and Learn: Unleashing Idle-Time Compute in Proactive Agents

While AI agents demonstrate remarkable capabilities in reasoning and tool use, they remain fundamentally reactive: they compute responses only after explicit user prompts. This paradigm ignores a critical opportunity: the idle time between interactions is largely wasted, leaving agents unable to prepare for future user needs. To bridge this gap, we introduce ProAct, a proactive agent architecture that leverages idle-time compute to anticipate and fulfill likely upcoming user needs. By analyzing evolving dialogue history together with persistent memory, ProAct predicts upcoming needs and iteratively acquires information, allowing the agent to resolve knowledge gaps and prepare evidence before the user initiates a query. To rigorously evaluate proactive capabilities, we also introduce ProActEval, a comprehensive benchmark comprising 200 scenarios across 40 domains, featuring predictable need chains and diverse user cognitive profiles. Empirical results demonstrate significant advantages over reactive baselines. ProAct accelerates task completion by reducing required turns by 14.8%, decreases user effort by 11.7%, and cuts hallucination rates by 28.1% on ProActEval. Furthermore, MemBench evaluations confirm that ProAct achieves state-of-the-art reflective accuracy, underscoring its sustained and robust performance.

cs.CL↗
arXiv2026

MedVol-R1: Reward-Driven Evidence Grounding for Volumetric Reasoning Segmentation

Volumetric Reasoning Segmentation (VRS) aims to segment a target region in a 3D medical scan from a free-form clinical query, where the referent is often implicit and requires both medical knowledge and volume-grounded reasoning. Existing methods typically rely on specialized segmentation tokens to connect language with mask decoding, but this coupling collapses the decision process into opaque latent representations, limiting interpretability and generalization to diverse narrative expressions. In this paper, we present MedVol-R1, a reinforcement learning-based framework for VRS that explicitly decouples evidence grounding from volumetric delineation: the LVLM grounds clinical reasoning to a verifiable 2D evidence anchor (key axial slice and 2D bounding boxes), which is then propagated into a coherent 3D mask by a frozen MedSAM2 module. We train MedVol-R1 with cold-start supervised fine-tuning followed by GRPO, guided by a multi-component reward that encourages informative evidence selection, accurate 2D spatial grounding, and cross-slice volumetric coherence, without requiring costly chain-of-thought annotations. Experiments on CT-ORG, AbdomenCT-1K, and KiTS23 from the M3D-Seg benchmark demonstrate that MedVol-R1 consistently outperforms strong baselines and achieves state-of-the-art performance, with reinforcement learning providing clear gains over pure supervised fine-tuning.

cs.CV↗
arXiv2026

Backward Uniqueness for Coupled Ultraparabolic Operators and an Application to Jerk-Driven Control Models

We prove backward uniqueness for a class of ultraparabolic operators with coupled linear drift. The main difficulty is that the Fourier transform in the degenerate variables turns the coupled drift into a transport operator in the dual frequency variables, so the classical Littlewood--Paley Carleman argument does not apply directly. We overcome this by introducing an invariant frequency variable and establishing a frequency-localized Carleman estimate adapted to the transport structure. The result gives a partial answer to the question of W. Wang and L. Zhang $\left[ \emph {Methods Appl. Anal.}, \ 20 \ (1) \ (2013) \ 79-88 \right]$ for constant coupled drift, with diffusion and lower-order coefficients depending on time and the diffusive variables. As an application, for a jerk-driven control model, we prove backward uniqueness for the equation describing the position, velocity, acceleration, or jerk error: under bounded lower-order coefficients, zero final error in $L^2$ implies zero error at all earlier times.

math.AP↗
arXiv2026

Deep Learning-Enabled Prediction of Geoeffective CMEs Using SOHO and SDO Observations

Understanding and forecasting the geoeffectiveness of a coronal mass ejection (CME) is crucial for protecting infrastructure in the near-Earth space environment and on Earth. In this study, we present a novel fusion model to forecast the geoeffectiveness of CME events. Our model combines convolutional neural networks for feature learning and a prediction network for feature fusion and event classification. The model is trained by observations from instruments including the Large Angle Spectroscopic Coronagraph (LASCO) on board the Solar and Heliospheric Observatory (SOHO) and the Atmospheric Imaging Assembly (AIA) and Helioseismic and Magnetic Imager (HMI) on board the Solar Dynamics Observatory (SDO). The trained model is then used to predict whether an Earth-reaching CME will cause a geomagnetic storm and/or the probability that the CME will cause such a storm. Experimental results based on a five-fold cross validation scheme demonstrate the good performance of our fusion model, achieving a mean true skill statistic (TSS) score of 0.703 when the model is used as a deterministic prediction tool, and a mean Brier score of 0.095 when the model is used as a probabilistic forecasting tool, where a TSS score of 1 or a Brier score of 0 indicates perfect performance. This work contributes to forecasting the causal relationship between Earth-directed CMEs and geomagnetic storms in solar-terrestrial interactions.

astro-ph.SR↗
arXiv2026

SkillEvolBench: Benchmarking the Evolution from Episodic Experience to Procedural Skills

Large language model (LLM) agents accumulate rich episodic trajectories while solving real-world tasks, but it remains unclear whether such experience can be distilled into reusable procedural skills. We introduce SkillEvolBench, a diagnostic benchmark for evaluating this step from experience reuse to skill formation. It contains 180 tasks across six real-world agent environments, organized into role-conditioned task families with shared latent procedures. Agents learn from acquisition tasks, update an external skill library using compacted trajectories and verifier feedback, and then face frozen deployment tasks testing context shift, adversarial shortcuts, and composition. By comparing self-generated and curated-start skill evolution against no-skill and raw-trajectory controls, SkillEvolBench separates procedural abstraction from base capability, curated prior knowledge, and direct reuse of episodic traces. Across ten model configurations and three agent harnesses, we find that current agents often adapt locally but rarely form robust reusable skills. Skill-based conditions can improve acquisition or replay, and individual models sometimes gain on specific deployment axes, but these gains are unstable under frozen deployment. Raw-trajectory reuse frequently outperforms distilled skills, suggesting that current abstraction procedures discard contextual and procedural cues that remain useful for future tasks. Capacity and cost analyses further show that writing more skills or larger Tier-3 resource libraries is not sufficient: additional updates can improve coverage while introducing episode-specific drift and procedural clutter. These findings position SkillEvolBench as a testbed for measuring when one-off experience becomes durable procedural knowledge rather than task-local memory.

cs.AI↗
arXiv2026

PIVOT: Bridging Planning and Execution in LLM Agents via Trajectory Refinement

Large language model (LLM)-based agents frequently generate seemingly coherent plans that fail upon execution due to infeasible actions, constraint violations, and compounding errors over extended horizons. PIVOT (Plan-Inspect-eVOlve Trajectories) addresses this plan-execution misalignment through a self-supervised framework that treats trajectories as optimizable objects iteratively refined via environment interaction. The framework comprises four stages: PLAN generates candidate trajectories; INSPECT executes them and computes structured losses with textual gradients encoding plan-execution discrepancies; EVOLVE applies these signals to produce improved trajectories; and VERIFY performs a final global check against task constraints. A monotonic acceptance process ensures a non-decreasing solution quality. Empirical evaluations on DeepPlanning and GAIA demonstrate state-of-the-art performance: with human-in-the-loop (HITL) feedback, PIVOT establishes a strong upper bound up to 94% relative improvement in constraint satisfaction, while its fully autonomous variant retains substantial gains, showing that the core trajectory-refinement mechanism remains effective without external supervision. At the same time, PIVOT remains computationally efficient, requiring up to 3x to 5x fewer tokens than competing refinement methods. These findings establish that (self- or human-supervised) feedback-based trajectory optimization is a principled methodology for mitigating plan-execution gaps in autonomous agent systems.

cs.AI↗
arXiv2026

EngiAgent: Fully Connected Coordination of LLM Agents for Solving Open-ended Engineering Problems with Feasible Solutions

Engineering problem solving is central to real-world decision-making, requiring mathematical formulations that not only represent complex problems but also produce feasible solutions under data and physical constraints. Unlike mathematical problem solving, which operates on predefined formulations, engineering tasks demand open-ended analysis, feasibility-driven modeling, and iterative refinement. Although large language models (LLMs) have shown strong capabilities in reasoning and code generation, they often fail to ensure feasibility, which limits their applicability to engineering problem solving. To address this challenge, we propose EngiAgent, a multi-agent system with a fully connected coordinator that simulates expert workflows through specialized agents for problem analysis, modeling, verification, solving, and solution evaluation. The fully connected coordinator enables flexible feedback routing, overcoming the rigidity of prior pipeline-based reflection methods and ensuring feasibility at every stage of the process. This design not only improves robustness to diverse failure cases such as data extraction errors, constraint inconsistencies, and solver failures, but also enhances the overall quality of problem solving. Empirical results across four representative domains demonstrate that EngiAgent achieves substantial improvements in feasibility compared to prior approaches, establishing a new paradigm for feasibility-oriented engineering problem solving with LLMs. Our source code and data are available at https://github.com/AI4Engi/EngiAgent.

cs.AI↗
arXiv2026

EngiBench: A Benchmark for Evaluating Large Language Models on Engineering Problem Solving

Large language models (LLMs) have shown strong performance on mathematical reasoning under well-defined conditions. However, real-world engineering problems involve uncertainty, context, and open-ended settings that extend beyond symbolic computation. Existing benchmarks largely focus on well-defined or abstract reasoning and therefore fail to capture these complexities. We introduce EngiBench, a hierarchical benchmark designed to evaluate LLMs on solving engineering problems. It spans three levels of increasing difficulty (foundational knowledge retrieval, contextual reasoning, and open-ended modeling) and covers diverse engineering subfields. To facilitate a deeper understanding of model performance, we systematically rewrite each problem into three controlled variants (perturbed, knowledge-enhanced, and math abstraction), enabling us to separately evaluate the model's robustness, domain-specific knowledge, and mathematical reasoning abilities. Experimental results show clear performance stratification across difficulty levels: model accuracy declines with task complexity, degrades under minor perturbations, and remains substantially below human performance on high-level engineering tasks. These findings reveal that current LLMs still lack the high-level reasoning needed for real-world engineering, highlighting the need for future models with deeper and more reliable problem-solving capabilities. Our source code and data are available at https://github.com/AI4Engi/EngiBench.

cs.AI↗
arXiv2026

Revealing a Systematic High-latitude Current Sheet at Jupiter

Based on models derived from Earth's magnetotail, other planets with dipole magnetic fields, including Mercury, Jupiter, and Saturn, were expected to possess similar magnetotail configurations. In this traditional picture, the majority of plasma is confined near the magnetic equator within a plasma sheet (or plasma disc), whereas higher-latitude regions feature strong magnetic fields that are open to the solar wind, forming magnetospheric lobes. However, auroral observations and recent simulations have shown that Jupiter's magnetic topology differs markedly from this picture, particularly in its high-latitude regions where magnetic field lines are predominantly closed. This discrepancy calls for a re-examination of high-latitude magnetospheric structure at Jupiter. Here, using Juno measurements acquired between 2016 and 2022, we show that Jupiter's nightside high latitudes host a persistent current-sheet-like structure above about 40 degrees magnetic latitude near midnight. This structure contains internally sourced oxygen and sulfur ions and exhibits azimuthal magnetic signatures opposite to the bend-back of the equatorial current sheet. These findings indicate that the canonical picture of planetary magnetotail architecture requires revision. Our results provide new insight into the architecture of rapidly rotating magnetospheres and offer a framework for interpreting magnetospheric structures at exoplanets.

astro-ph.EP↗
arXiv2026

Co-Director: Agentic Generative Video Storytelling

While diffusion models generate high-fidelity video clips, transforming them into coherent storytelling engines remains challenging. Current agentic pipelines automate this via chained modules but suffer from semantic drift and cascading failures due to independent, handcrafted prompting. We present Co-Director, a hierarchical multi-agent framework formalizing video storytelling as a global optimization problem. To ensure semantic coherence, we introduce hierarchical parameterization: a multi-armed bandit globally identifies promising creative directions, while a local multimodal self-refinement loop mitigates identity drift and ensures sequence-level consistency. This balances the exploration of novel narrative strategies with the exploitation of effective creative configurations. For evaluation, we introduce GenAD-Bench, a 400-scenario dataset of fictional products for personalized advertising. Experiments demonstrate that Co-Director significantly outperforms state-of-the-art baselines, offering a principled approach that seamlessly generalizes to broader cinematic narratives. Project Page: https://co-director-agent.github.io/

cs.AI↗
arXiv2026

Discover and Prove: An Open-source Agentic Framework for Hard Mode Automated Theorem Proving in Lean 4

Most ATP benchmarks embed the final answer within the formal statement -- a convention we call "Easy Mode" -- a design that simplifies the task relative to what human competitors face and may lead to optimistic estimates of model capability. We call the stricter, more realistic setting "Hard Mode": the system must independently discover the answer before constructing a formal proof. To enable Hard Mode research, we make two contributions. First, we release MiniF2F-Hard and FIMO-Hard, expert-reannotated Hard Mode variants of two widely-used ATP benchmarks. Second, we introduce Discover And Prove (DAP), an agentic framework that uses LLM natural-language reasoning with explicit self-reflection to discover answers, then rewrites Hard Mode statements into Easy Mode ones for existing ATP provers. DAP sets the state of the art: on CombiBench it raises solved problems from 7 (previous SOTA, Pass@16) to 10; on PutnamBench it is the first system to formally prove 36 theorems in Hard Mode -- while simultaneously revealing that state-of-the-art LLMs exceed 80% answer accuracy on the same problems where formal provers manage under 10%, exposing a substantial gap that Hard Mode benchmarks are uniquely suited to measure.

cs.AI↗