arXiv subjects

Yuwei Zhang

Publications and source records attributed to Yuwei Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Learning with Rare Success but Rich Feedback via Reflection-Enhanced Self-Distillation

Enabling Large Language Models (LLMs) to continuously improve from environmental interactions is a central challenge in post-training. While on-policy self-distillation offers a promising paradigm, existing methods predominantly treat environmental feedback as a passive conditioning signal. Consequently, they heavily rely on successful demonstrations and struggle to learn in rare-success regimes. To bridge this gap, we introduce Reflection-Enhanced Self-Distillation (RESD), a framework that transforms raw failure feedback into an active source of corrective supervision. Instead of passively appending feedback, RESD interprets failed trajectories by generating retrospective reflections to diagnose local errors, and curates a persistent global playbook to preserve reusable lessons across training steps. The enriched context enables the self-teacher to provide actionable token-level supervision even in the absence of successful rollouts. Empirical evaluations on multiple continual learning tasks demonstrate that RESD substantially outperforms standard self-distillation baselines. Furthermore, RESD achieves significantly faster early-stage improvement than GRPO with $8\times$ samples using only a single rollout per prompt, highlighting its superior interaction efficiency.

cs.LG↗

arXiv2026

MaPPO: Maximum a Posteriori Preference Optimization with Prior Knowledge

As the era of large language models (LLMs) unfolds, Preference Optimization (PO) methods have become a central approach to aligning LLMs with human preferences and improving performance. We propose Maximum a Posteriori Preference Optimization (MaPPO), a methodology for learning from preferences that explicitly incorporates prior reward knowledge into the optimization objective. Building on the paradigm employed by Direct Preference Optimization (DPO) and its variants of treating preference learning as a Maximum Likelihood Estimation (MLE) problem, MaPPO integrates prior reward estimates into a principled Maximum a Posteriori (MaP) objective. This not only generalizes DPO and its variants, but also enhances alignment by mitigating the oversimplified binary classification of responses. Additionally, MaPPO introduces no additional hyperparameters, and supports preference optimization in both offline and online settings. In addition, MaPPO can be used as a plugin for DPO variants, including widely used SimPO, IPO and CPO, and produce consistent improvements. Extensive empirical evaluations of different model sizes and model series on three standard benchmarks (MT-Bench, AlpacaEval 2.0, and Arena-Hard) demonstrate consistent improvements in alignment performance without sacrificing computational efficiency.

cs.LG↗

arXiv2026

Alternating Reinforcement Learning with Contextual Rubric Rewards: Beyond the Scalarization Strategy

Reinforcement Learning with Rubric Rewards (RLRR) is a framework that extends conventional reinforcement learning from human feedback (RLHF) and verifiable rewards (RLVR) by replacing scalar preference signals with structured, multi-dimensional, contextual rubric-based evaluations. However, existing approaches in RLRR are limited to linearly compressing vector rewards into a scalar reward with a fixed weightings, which is sensitive to artificial score design and fails to capture correlations among reward dimensions. To overcome the limitations of reward aggregation, this work proposes Alternating Reinforcement Learning with Rubric Rewards (ARL-RR), a framework that eliminates the need for a fixed scalarization by optimizing one semantic rubric meta-class at a time. Theoretically, we show that reward aggregation induces a variance contraction effect, which helps explain the performance gains. We further introduce a lightweight, search-based adaptation procedure that selects the next meta-class dynamically based on task performance, enabling the policy to emphasize critical objectives and thereby improve the model performance. Empirically, our experiments on the HealthBench dataset with experts annotations demonstrate that ARL-RR uniformly outperforms scalarized methods in both model performance and training efficiency across different model scales (1.7B, 4B, 8B, and 14B).

cs.LG↗

arXiv2026

RAMoEA-QA: Hierarchical Specialization for Robust Respiratory Audio Question Answering

Conversational generative AI is increasingly explored in healthcare, where models must integrate heterogeneous patient signals and support diverse interaction styles while producing clinically meaningful outputs. In respiratory care, non-invasive audio recordings captured with sensing devices offer a scalable route to screening and longitudinal monitoring, but heterogeneity is particularly acute: recordings vary across devices, environments, and acquisition protocols, and queries may vary in intent, answer format, and prediction objective. Existing biomedical audio-language question answering systems for respiratory assessment are starting to emerge, but they are typically built as single-path models, processing all inputs through the same acoustic and language pathway despite variation in recording conditions and query types. They are also usually evaluated in relatively limited settings, leaving open their robustness under realistic distribution shifts, including changes in acquisition domains, modality, and clinical task. To address this gap, we introduce RAMoEA-QA, the first RA QA model designed to support input-dependent specialization across heterogeneous recordings and query types within a unified hierarchical two-stage framework. We study this design in a unified RA QA setting spanning clinical and self-recorded, multi-device acquisition settings, question formats, and both discrete and continuous targets. Across in-domain and controlled-shift evaluations, RAMoEA-QA improves over matched monolithic baselines and routing controls, reaching 0.72 in in-domain test accuracy (vs. 0.61 and 0.67 for single-path baselines) on discriminative tasks, while also achieving the best regression performance and stronger average transfer under dataset, modality, and task shifts, including gains of up to 23 percentage points in accuracy on the COPD modality-shift setting.

cs.SD↗

arXiv2026

Compositional Complexity-Induced Ultralow Friction in Medium-Entropy MXenes

Two-dimensional MXenes are promising solid lubricants, but the roles of compositional complexity and surface chemistry in governing interfacial friction remain unclear. Here, we systematically investigate the adhesion and friction behaviors of medium-entropy (ME) MXenes, TiVNbMoC3 and TiVCrMoC3, and compare them with conventional titanium carbide MXenes, Ti2C and Ti3C2, using a SiO2 colloidal atomic force microscopy probe. Thermal annealing at 200 C converts OH surface terminations to O terminations, leading to pronounced reductions in adhesion energy and friction force across all MXenes studied. ME MXenes exhibit larger adhesion reductions because of their higher initial OH contents and more extensive OH-to-O conversion. In addition, their intrinsically higher out-of-plane bending stiffness suppresses energy dissipation during sliding, enabling ultralow friction. Notably, superlubricity is achieved in ME MXenes, with annealed TiVCrMoC3 exhibiting a coefficient of friction as low as 0.0022, outperforming graphene, MoSe2, and other MXenes evaluated using the same experimental approach. These findings identify compositional complexity as a powerful strategy for engineering MXenes with exceptional tribological performance and establish ME MXenes as a new class of solid lubricants.

cond-mat.mtrl-sci↗

arXiv2026

Wearable Foundation Models Should Go Beyond Static Encoders

Wearable foundation models (WFMs), trained on large volumes of data collected by affordable, always-on devices, have demonstrated strong performance on short-term, well-defined health monitoring tasks, including activity recognition, fitness tracking, and cardiovascular signal assessment. However, most existing WFMs primarily map short temporal windows to predefined labels via static encoders, emphasizing retrospective prediction rather than reasoning over evolving personal history, context, and future risk trajectories. As a result, they are poorly suited for modeling chronic, progressive, or episodic health conditions that unfold over weeks, months or years. Hence, we argue that WFMs must move beyond static encoders and be explicitly designed for longitudinal, anticipatory health reasoning. We identify three foundational shifts required to enable this transition: (1) Structurally rich data, which goes beyond isolated datasets or outcome-conditioned collection to integrated multimodal, long-term personal trajectories, and contextual metadata, ideally supported by open and interoperable data ecosystems; (2) Longitudinal-aware multimodal modeling, which prioritizes long-context inference, temporal abstraction, and personalization over cross-sectional or population-level prediction; and (3) Agentic inference systems, which move beyond static prediction to support planning, decision-making, and clinically grounded intervention under uncertainty. Together, these shifts reframe wearable health monitoring from retrospective signal interpretation toward continuous, anticipatory, and human-aligned health support.

cs.LG↗

arXiv2026

Large room temperature anomalous Nernst effect coupled with topological Nernst effect from incommensurate spin structure in a Kagome antiferromagnet

Kagome magnets exhibit a range of novel and nontrivial topological properties due to the strong interplay between topology and magnetism, which also extends to their thermoelectric applications. Recent advances in the study of magnetic topological materials have highlighted their intriguing anomalous Hall and thermoelectric effects, arising primarily from large intrinsic Berry curvature. Here, we report observation of a large room-temperature (RT) anomalous Nernst effects (ANE) of S_xy^A ~ 1.3 μV K^(-1) in the kagome antiferromagnet (AFM) ErMn6Sn6, which is comparable to the largest signals observed in known magnetic materials. Surprisingly, we further found that a significant topological Nernst signal at RT and peaking a maximum of approximately 0.2 μV K^(-1) at 180 K, exactly coupling with ANE in the spiral AFM state, originates from the real-space nonzero spin chirality caused by incommensurate spin structure. This study demonstrates a potential room-temperature thermoelectric application platform based on Nernst effect, and provides insights for discovering significant anomalous and topological transverse transport effects in the incommensurate AFM system.

cond-mat.str-el↗

arXiv2026

RISE: Rule-Driven SQL Dialect Translation via Query Reduction

Translating SQL dialects across different relational database management systems (RDBMSs) is crucial for migrating RDBMS-based applications to the cloud. Traditional SQL dialect translation tools rely on manually-crafted rules, necessitating significant manual effort to support new RDBMSs and dialects. Although large language models (LLMs) can assist in translating SQL dialects, they often struggle with lengthy and complex SQL queries. In this paper, we propose RISE, a novel LLM-based SQL dialect translation approach that can accurately handle lengthy and complex SQL queries. Given a complex source query $Q_c$ that contains a SQL dialect $d$, we first employ a dialect-aware query reduction technique to derive a simplified query $Q_{s}$ by removing $d$-irrelevant SQL elements from $Q_c$. Subsequently, we utilize LLMs to translate $Q_{s}$ into $Q_{s^{'}}$, and automatically extract the translation rule $r_d$ for dialect $d$ based on the relationship between $Q_{s}$ and $Q_{s^{'}}$. By applying $r_d$ to $Q_c$, we can effectively translate the dialect $d$ within $Q_c$, thereby bypassing the complexity of the source query $Q_c$. We evaluate RISE on two real-world benchmarks, i.e., TPC-DS and SQLProcBench, comparing its performance against both the traditional rule-based tools and the LLM-based approaches with respect to translation accuracy. RISE achieves accuracies of 97.98% on TPC-DS and 100% on SQLProcBench, outperforming the baselines by an average improvement of 24.62% and 238.41%, respectively.

cs.DB↗

arXiv2026

NC2C: Automated Convexification of Generic Non-Convex Optimization Problems

Non-convex optimization problems are pervasive across mathematical programming, engineering design, and scientific computing, often posing intractable challenges for traditional solvers due to their complex objective functions and constrained landscapes. To address the inefficiency of manual convexification and the over-reliance on expert knowledge, we propose NC2C, an LLM-based end-to-end automated framework designed to transform generic non-convex optimization problems into solvable convex forms using large language models. NC2C leverages LLMs' mathematical reasoning capabilities to autonomously detect non-convex components, select optimal convexification strategies, and generate rigorous convex equivalents. The framework integrates symbolic reasoning, adaptive transformation techniques, and iterative validation, equipped with error correction loops and feasibility domain correction mechanisms to ensure the robustness and validity of transformed problems. Experimental results on a diverse dataset of 100 generic non-convex problems demonstrate that NC2C achieves an 89.3\% execution rate and a 76\% success rate in producing feasible, high-quality convex transformations. This outperforms baseline methods by a significant margin, highlighting NC2C's ability to leverage LLMs for automated non-convex to convex transformation, reduce expert dependency, and enable efficient deployment of convex solvers for previously intractable optimization tasks.

cs.CL↗

arXiv2025

Seedream 4.0: Toward Next-generation Multimodal Image Generation

We introduce Seedream 4.0, an efficient and high-performance multimodal image generation system that unifies text-to-image (T2I) synthesis, image editing, and multi-image composition within a single framework. We develop a highly efficient diffusion transformer with a powerful VAE which also can reduce the number of image tokens considerably. This allows for efficient training of our model, and enables it to fast generate native high-resolution images (e.g., 1K-4K). Seedream 4.0 is pretrained on billions of text-image pairs spanning diverse taxonomies and knowledge-centric concepts. Comprehensive data collection across hundreds of vertical scenarios, coupled with optimized strategies, ensures stable and large-scale training, with strong generalization. By incorporating a carefully fine-tuned VLM model, we perform multi-modal post-training for training both T2I and image editing tasks jointly. For inference acceleration, we integrate adversarial distillation, distribution matching, and quantization, as well as speculative decoding. It achieves an inference time of up to 1.8 seconds for generating a 2K image (without a LLM/VLM as PE model). Comprehensive evaluations reveal that Seedream 4.0 can achieve state-of-the-art results on both T2I and multimodal image editing. In particular, it demonstrates exceptional multimodal capabilities in complex tasks, including precise image editing and in-context reasoning, and also allows for multi-image reference, and can generate multiple output images. This extends traditional T2I systems into an more interactive and multidimensional creative tool, pushing the boundary of generative AI for both creativity and professional applications. We further scale our model and data as Seedream 4.5. Seedream 4.0 and Seedream 4.5 are accessible on Volcano Engine https://www.volcengine.com/experience/ark?launch=seedream.

cs.CV↗

arXiv2025

GraphCodeAgent: Dual Graph-Guided LLM Agent for Retrieval-Augmented Repo-Level Code Generation

Writing code requires significant time and effort in software development. To automate this process, researchers have made substantial progress for code generation. Recently, large language models (LLMs) have demonstrated remarkable proficiency in function-level code generation, yet their performance significantly degrades in the real-world software development process, where coding tasks are deeply embedded within specific repository contexts. Existing studies attempt to use retrieval-augmented code generation (RACG) approaches to mitigate this demand. However, there is a gap between natural language (NL) requirements and programming implementations. This results in the failure to retrieve the relevant code of these fine-grained subtasks. To address this challenge, we propose GraphCodeAgent, a dual graph-guided LLM agent for retrieval-augmented repo-level code generation, bridging the gap between NL requirements and programming implementations. Our approach constructs two interconnected graphs: a Requirement Graph (RG) to model requirement relations of code snippets within the repository, as well as the relations between the target requirement and the requirements of these code snippets, and a Structural-Semantic Code Graph (SSCG) to capture the repository's intricate code dependencies. Guided by this, an LLM-powered agent performs multi-hop reasoning to systematically retrieve all context code snippets, including implicit and explicit code snippets, even if they are not explicitly expressed in requirements. We evaluated GraphCodeAgent on three advanced LLMs with the two widely-used repo-level code generation benchmarks DevEval and CoderEval. Extensive experiment results show that GraphCodeAgent significantly outperforms state-of-the-art baselines.

cs.SE↗

arXiv2025

Scale-bridging dislocation plasticity in MgO at room temperature

Dislocations in ceramics have recently gained renewed research interest, in contrast to the traditional belief that ceramics are inherently brittle. Understanding dislocation mechanics in representative oxides is beneficial for effective dislocation engineering. Here, we use MgO single crystals with mechanically seeded dislocation densities from about 10 to the power of 12 to about 10 to the power of 15 per square meter to investigate the mechanical behavior such as yield and fracture. Micro-pillar compression tests reveal a dislocation density dependent yield strength, mediated by the varying dominating dislocation mechanisms from nucleation to multiplication/motion. In situ TEM compression measurements highlight the dislocation-seeded samples can achieve a much-improved compressive plastic strain beyond about 70%, with a high yield strength of about 2.35 GPa (diameter of about 400 nm), indicating size effect. Complementary bulk compression tests, along with digital image correlation (DIC), demonstrate a consistent dislocation-mediated deformation and a notable size effect, with bulk samples exhibiting much reduced yield strength (about 120 MPa) compared to the nano-/micro-pillars. Using three-dimensional Discrete Dislocation Dynamics (3D-DDD) simulation, we further qualitatively analyze the collective dislocation activities (slip events) and work hardening during compression. This study provides new insights into dislocation-mediated plasticity in MgO, across different length scales, by systematically tuning dislocation density.

cond-mat.mtrl-sci↗

arXiv2025

RADAR: Benchmarking Language Models on Imperfect Tabular Data

Language models (LMs) are increasingly being deployed to perform autonomous data analyses. However, their data awareness -- the ability to recognize, reason over, and appropriately handle data artifacts such as missing values, outliers, and logical inconsistencies -- remains underexplored. These artifacts are especially common in real-world tabular data and, if mishandled, can significantly compromise the validity of analytical conclusions. To address this gap, we present RADAR, a benchmark for systematically evaluating data-aware reasoning on tabular data. We develop a framework to simulate data artifacts via programmatic perturbations to enable targeted evaluation of model behavior. RADAR comprises 2980 table query pairs, grounded in real-world data spanning 9 domains and 5 data artifact types. In addition to evaluating artifact handling, RADAR systematically varies table size to study how reasoning performance holds when increasing table size. Our evaluation reveals that, despite decent performance on tables without data artifacts, frontier models degrade significantly when data artifacts are introduced, exposing critical gaps in their capacity for robust, data-aware analysis. Designed to be flexible and extensible, RADAR supports diverse perturbation types and controllable table sizes, offering a valuable resource for advancing tabular reasoning.

cs.DB↗

arXiv2025

The Anatomy of a Personal Health Agent

Health is a fundamental pillar of human wellness, and the rapid advancements in large language models (LLMs) have driven the development of a new generation of health agents. However, the application of health agents to fulfill the diverse needs of individuals in daily non-clinical settings is underexplored. In this work, we aim to build a comprehensive personal health agent that is able to reason about multimodal data from everyday consumer wellness devices and common personal health records, and provide personalized health recommendations. To understand end-users' needs when interacting with such an assistant, we conducted an in-depth analysis of web search and health forum queries, alongside qualitative insights from users and health experts gathered through a user-centered design process. Based on these findings, we identified three major categories of consumer health needs, each of which is supported by a specialist sub-agent: (1) a data science agent that analyzes personal time-series wearable and health record data, (2) a health domain expert agent that integrates users' health and contextual data to generate accurate, personalized insights, and (3) a health coach agent that synthesizes data insights, guiding users using a specified psychological strategy and tracking users' progress. Furthermore, we propose and develop the Personal Health Agent (PHA), a multi-agent framework that enables dynamic, personalized interactions to address individual health needs. To evaluate each sub-agent and the multi-agent system, we conducted automated and human evaluations across 10 benchmark tasks, involving more than 7,000 annotations and 1,100 hours of effort from health experts and end-users. Our work represents the most comprehensive evaluation of a health agent to date and establishes a strong foundation towards the futuristic vision of a personal health agent accessible to everyone.

cs.AI↗

arXiv2025

Anomalous Nernst Effect and Its Implications for Time-Reversal Symmetry Breaking in Kagome Metal ScV6Sn6

The nonmagnetic kagome metal ScV6Sn6 displays an unconventional charge order (CO) accompanied by signatures of an anomalous Hall effect, hidden magnetism, and multiple lattice instabilities. In this study, we report the observation of unconventional anomalous thermoelectric properties. Notably, unexpected anomalous transverse Nernst signals reach a peak value of ~4 μV/K near the TCDW ~92 K in ScV6Sn6, and these signals persist in the charge-ordered state as the temperature decreases to 10 K. Furthermore, both thermopower and thermal conductivity exhibit significant changes under magnetic fields, even in the nonmagnetic ground state. These observations strongly suggest the emergence of time-reversal symmetry breaking in ScV6Sn6, as supported by muon spin relaxation (μSR) measurements. While hidden magnetism represents the most plausible origin, alternative mechanisms involving orbital currents and chiral charge order remain possible.

cond-mat.str-el↗

arXiv2025

Tool-Planner: Task Planning with Clusters across Multiple Tools

Large language models (LLMs) have demonstrated exceptional reasoning capabilities, enabling them to solve various complex problems. Recently, this ability has been applied to the paradigm of tool learning. Tool learning involves providing examples of tool usage and their corresponding functions, allowing LLMs to formulate plans and demonstrate the process of invoking and executing each tool. LLMs can address tasks that they cannot complete independently, thereby enhancing their potential across different tasks. However, this approach faces two key challenges. First, redundant error correction leads to unstable planning and long execution time. Additionally, designing a correct plan among multiple tools is also a challenge in tool learning. To address these issues, we propose Tool-Planner, a task-processing framework based on toolkits. Tool-Planner groups tools based on the API functions with the same function into a toolkit and allows LLMs to implement planning across the various toolkits. When a tool error occurs, the language model can reselect and adjust tools based on the toolkit. Experiments show that our approach demonstrates a high pass and win rate across different datasets and optimizes the planning scheme for tool learning in models such as GPT-4 and Claude 3, showcasing the potential of our method. Our code is public at https://github.com/OceannTwT/Tool-Planner

cs.AI↗

arXiv2025

Electrochemically-driven formation of Intermetallic Cu3ZnLi2 alters Li-transport in nanostructured bimetallic battery anode

The role of Li-based batteries in the electrification of society cannot be understated, however their operational lifetime is often limited by the formation of dendrites, i.e. the localised deposition of Li that can cause shorts between the two electrodes leading to the failure of the battery. Nanocrystalline bimetallic current collectors can be used for anode-free Li-metal batteries, with improved Li plating and limited or suppressed formation of dendrites. Here, we demonstrate that the microstructure of an alpha-Brass current collector, Cu 63% Zn 37%, used in an anode-free Li-metal battery evolves during cycling. It initially had a nanocrystalline deformation layer approximately 80 nm in thickness after polishing. After 100 cycles, the initial deformed brass layer was partially converted to a ternary Laves phase Cu3ZnLi2 within a nanocrystalline brass matrix that grew to 200 - 250 nm in thickness. Upon Li stripping, the phase partially decomposes electrochemically, but what remains can sequester Li thus forming "dead Li" thereby contributing to capacity loss. We propose a mechanism for the microstructural evolution including dynamic recrystallization and phase formation. Since this ternary Laves phase emerges during electrochemical cycling alone, binary alloy current collectors must be assessed for metastable ternary phase formation under different cycling conditions to either stabilize and exploit such phases or electrochemically fully strip them.

cond-mat.mtrl-sci↗

arXiv2025

Vacuum Dealloyed Brass as Li-Metal Battery Current Collector: Effect of Zinc and Porosity

"Anode-free" lithium-metal batteries promise significantly higher energy density than conventional graphite-based lithium-ion batteries; however, lithium dendrite growth can lead to internal short circuits with associated safety risks. While porous current collectors can suppress dendrite growth, optimal porosity and composition remain unknown. Here, we show that the temperature during vapor phase dealloying (VPD) of alpha-brass (Cu63Zn37) controls the surface Zn concentration, decreasing from 8 percent to below 1 percent from 500 to 800 degrees C. The surface composition is controlled by the temperature-dependent diffusion. A battery cell maintains greater than 90 percent Coulombic efficiency (CE) over 100 cycles when the Zn content is the lowest, whereas the higher-Zn samples degraded to approximately 70 percent CE. The difference in surface composition has hence dramatic effects on battery performance, and our results demonstrate how precise compositional control enables stable lithium-metal battery operation, establishing about 1 atomic percent surface Zn as optimal for preventing capacity fading and uniform lithium plating, while establishing predictive relationships between processing temperature and surface composition. This work provides design rules for multifunctional current collectors and demonstrates scalable VPD production for next-generation batteries.

cond-mat.mtrl-sci↗