arXiv subjects

Yi Zhang

Publications and source records attributed to Yi Zhang.

At least 181 records · Page 10Linked to original sources

arXiv2026

The FAST Framework: Developing a Data-Efficient Machine Learning Potential to Decode Superionic Transition-Induced Thermophysical and Kinetic Anomalies in UO2 under Extreme Conditions

Uranium dioxide ($UO_2$) serves as the predominant nuclear fuel globally. Despite its widespread application, evaluating its mechanical, thermophysical, and species transport behaviors under extreme accident scenarios remains a formidable challenge for conventional experimental and computational methods. To address this, we develop a versatile machine learning interatomic potential (MLIP) for $UO_2$ by proposing an efficient training strategy, termed the "FAST" (Fine-tuning via Active-learning and Superionic-Targeting) framework. Our "FAST" framework integrates superionic transition-targeted sampling with active learning-enhanced exploration to efficiently construct a highly compact dataset comprising only 500 configurations for fine-tuning a foundation model. By rigorously accounting for the strong correlation of uranium 5f electrons and antiferromagnetic (AFM) ground state during DFT labeling, we train a robust DFT-level neuroevolution potential (NEP) for $UO_2$. We demonstrate that this NEP exhibits superior predictive capability for various physical properties, encompassing mechanical, defect, thermophysical, and ionic diffusion over an extensive temperature range. Moreover, this NEP accurately captures the anomalous thermophysical and kinetic behaviors triggered by superionic transition. Specifically, it reproduces both the $λ$-peak in linear thermal expansion coefficient (LTEC) and "non-Arrhenius" anionic diffusion. Crucially, NEP-based simulations elucidate the microscopic origins underlying these anomalies: the pre-melting of oxygen sublattice and resultant kinetic decoupling between U and O ions.

cond-mat.mtrl-sci↗

arXiv2026

Galactic Diffuse Gamma-Ray and Neutrino Emission from Cosmic-Ray Interactions in Stellar Atmospheres

The Galactic diffuse gamma-ray emission is conventionally modeled as the product of cosmic-ray interactions with the interstellar medium. However, the cumulative contribution of stellar atmospheres acting as hadronic interaction targets remains an unexplored multi-messenger background. In this work, we present the first systematic evaluation of this stellar diffuse emission by coupling MESA stellar evolution profiles and magnetic-field-modulated cosmic-ray transport with a 3D Galactic population synthesis framework. We find that the cumulative stellar contribution to the Galactic diffuse gamma-ray flux is negligible at 1 TeV, and the associated diffuse neutrino flux ($\sim 10^{-16}\;\mathrm{TeV\;cm^{-2}\;s^{-1}\;sr^{-1}}$) remains orders of magnitude below current IceCube limits. Nevertheless, at ultra-high energies ($>10\;\mathrm{TeV}$), this emission establishes an irreducible local background that overtakes the strongly attenuated extragalactic isotropic gamma-ray background. Our results demonstrate that the Galactic stellar ensemble is a strictly sub-dominant background, indicating that stellar subtraction templates are not required for identifying Galactic PeVatrons or constraining dark matter annihilation.

astro-ph.HE↗

arXiv2026

Beyond the LUMIR challenge: The pathway to foundational registration models

Medical image challenges have played a transformative role in advancing the field, catalyzing innovation and establishing new performance benchmarks. Image registration, a foundational task in neuroimaging, has similarly advanced through the Learn2Reg initiative. Building on this, we introduce the Large-scale Unsupervised Brain MRI Image Registration (LUMIR) challenge, a next-generation benchmark for unsupervised brain MRI registration. Previous challenges relied upon anatomical label maps, however LUMIR provides 4,014 unlabeled T1-weighted MRIs for training, encouraging biologically plausible deformation modeling through self-supervision. Evaluation includes 590 in-domain test subjects and extensive zero-shot tasks across disease populations, imaging protocols, and species. Deep learning methods consistently achieved state-of-the-art performance and produced anatomically plausible, diffeomorphic deformation fields. They outperformed several leading optimization-based methods and remained robust to most domain shifts. These findings highlight the growing maturity of deep learning in neuroimaging registration and its potential to serve as a foundation model for general-purpose medical image registration.

eess.IV↗

arXiv2026

NIM4-ASR: Towards Efficient, Robust, and Customizable Real-Time LLM-Based ASR

Integrating large language models (LLMs) into automatic speech recognition (ASR) has become a mainstream paradigm in recent years. Although existing LLM-based ASR models demonstrate impressive performance on public benchmarks, their training remains predominantly data-driven, leaving key practical challenges insufficiently addressed -- particularly limited downward scalability in resource-constrained deployments and hallucinations under acoustically challenging conditions. To address these issues, we present NIM4-ASR, a production-oriented LLM-based ASR framework optimized for both efficiency and robustness. Grounded in a principled delineation of functional roles between the encoder and the LLM, we redesign the multi-stage training paradigm to align each module with its intended capability boundary. Specifically, we reformulate the pre-training architecture and objective to mitigate the modality gap and improve parameter efficiency; introduce an iterative asynchronous SFT stage to preserve acoustic fidelity and constrain representation drift; and design an ASR-specialized reinforcement learning stage to further enhance recognition quality and robustness. We additionally incorporate a suite of production-oriented optimizations, including robustness under noisy and silent conditions, real-time streaming inference, and hotword customization via retrieval-augmented generation (RAG). Experiments show that NIM4-ASR achieves state-of-the-art performance on multiple public benchmarks with merely 2.3B parameters, while substantially outperforming larger-scale competitors on internal benchmarks -- particularly in entity-intensive real-world scenarios. NIM4-ASR further supports million-scale hotword customization via RAG with sub-millisecond retrieval latency, enabling efficient adaptation to emerging entities and personalized user requirements.

eess.AS↗

arXiv2026

Current World Models Lack a Persistent State Core

World models are increasingly regarded as a decisive step toward artificial general intelligence, yet modeling the physical world demands more than rendering convincing frames on demand: it requires an internal world state that keeps evolving over time, decoupled from observation, so that objects endure and events run to their conclusions whether or not a camera is watching, much as the moon holds to its orbit when no one is looking. This requirement is a blind spot of existing benchmarks, which reward surface properties such as fidelity, motion, and camera controllability while never asking whether a generated world keeps evolving once it is unobserved. We introduce \textbf{WRBench}, the first systematic diagnostic benchmark that treats camera motion as an intervention on observability and resolves evaluation into a human-calibrated chain that asks whether the camera executes the requested interaction, whether the scene stays continuous and identifiable while in view, and whether a returning target remains consistent with the event that was set in motion. Across 9{,}600 videos from 23 models spanning four control paradigms, one finding proves stubborn: current systems maintain the observed world as a tracking shot, resuming a returning target in the state at which it was abandoned rather than advancing the event while it went unseen. Because this failure recurs across control paradigms, model families, and increments of scale, robust world-state evolution does not follow from cleaner imagery, tighter control, richer geometric priors, or sheer parameter count We therefore argue that the stability of the physical state kernel and the consistency of worldlines under viewpoint intervention should become first-class objectives of world-model design, so that a world model captures how the world will unfold rather than how the next frame appears.

cs.CV↗

arXiv2026

Global bounded solutions for a class of generalized Hillen-Painter models near Couette flow in $\mathbb{R}^2$

We investigate the global well-posedness of a class of generalized Hillen--Painter systems -- specifically, supercritical volume-filling chemotaxis models -- in $\mathbb{R}^2$ under the influence of Couette flow. It is well established that, in the absence of fluid flow, solutions to this system may develop finite-time singularities (blow-up) for arbitrary initial cell mass. It is proved that the introduction of a Couette flow with sufficiently large amplitude guarantees the global existence of solutions for all initial masses. By employing a novel frequency decomposition technique, we successfully remove the mass threshold limitation presented in previous studies on the domain $\mathbb{T}\times\mathbb{R}$ (Wang et al., Commun. Contemp. Math.), thereby establishing global regularity in the whole space without any smallness assumptions.

math.AP↗

arXiv2026

Hallucination Detection and Correction in Medical VLMs via Counter-Evidence Verification

Vision-Language models (VLMs) reliability in medical diagnosis is challenged by trust-undermining hallucinations. Existing hallucination detection approaches mainly focus on identifying factual inconsistencies between generated text and reference data. While some studies analyze where models attend in images, they seldom verify whether such attention truly reflects the visual evidence supporting the generated text. To address this gap, we propose Co}unter-Evidence Verification (CoEV), a training-free plug-and-play framework that detects and corrects hallucinations through evidence-based factual consistency verification. CoEV performs bidirectional verification between textual assertions and visual evidence, testing whether each statement is supported by its corresponding evidence region, and assigns each statement into a four-quadrant diagnostic map capturing combinations of text factuality and visual grounding. CoEV detects hallucinated content and serves as a post hoc refinement tool, correcting hallucinations without retraining. Extensive experiments on four medical datasets show that CoEV combats hallucinations in VLMs.For hallucination detection, CoEV consistently outperforms existing methods, improving average PR-AUC and ROC-AUC by 3.0% and 3.9% absolute points respectively, with notable gains of up to 18.5% in specific VQA scenarios. For hallucination correction, it improves Micro-F1 by up to 12.5%, reduces hallucination rates by over 11.9% on medical report generation, and also boosts medical VQA accuracy. These results show that CoEV enables reliable detection and correction of hallucinations, providing clinicians with dependable, evidence-based cues for diagnosis. Code will be released upon acceptance.

cs.CV↗

arXiv2026

General duality relations for hypergeometric and basic hypergeometric series

Duality relations for hypergeometric functions have reappeared as an active research topic several times, with the first instances tracing back to Euler and Gauss and the latest burst of activity occurring between 2015 and 2023. In this paper we present a common generalization of all relations of this type found in the existing literature both for hypergeometric and for $q$-hypergeometric functions. We cover both Gauss type and confluent generalized hypergeometric functions and their $q$-analogues. Our results entail a number of corollaries including multi-term relations for hypergeometric and $q$-hypergeometric series at a fixed argument.

math.CA↗

arXiv2026

Beyond Citations: Comparing Scholarly, Policy, and Patent Impact Across the FT50 Journals

The Financial Times 50 (FT50) journal list shapes hiring, promotion, accreditation, and research evaluation across business schools worldwide. Yet journals on the list are typically treated as if they represent a homogeneous tier of excellence. We test this assumption by comparing 53 FT50 and recently removed journals across three distinct impact channels: scholarly influence (field-weighted citations and visibility), policy uptake, and technological reach through patent citations. Using a panel of more than 60,000 publications from 2005 to 2019, we find striking heterogeneity hidden beneath the binary FT50 label. Elite economics journals dominate policy influence, information systems and marketing journals lead technological impact, while many highly cited management journals exhibit limited reach beyond academia. Citation, policy, and patent indicators behave as largely independent dimensions of impact, with a citation-only ranking correlating only moderately with a multidimensional ranking. Nearly half of all journals change quartile once policy and patent indicators are incorporated, demonstrating that assessments based solely on scholarly citations overlook important dimensions of research influence. While the FT50 remains widely used as a binary classification of journal quality, our results reveal a substantial within-list impact spectrum and show that journal rankings are highly sensitive to how impact is defined and measured.

cs.DL↗

arXiv2026

Psyzkaller: Learning from Historical and On-the-Fly Execution Data for Smarter Seed Generation in OS kernel Fuzzing

OS Kernel fuzzers such as Syzkaller often struggle to generate syscall sequences that respect intrinsic Syscall Dependency Relations (SDRs), resulting in seeds that either violate kernel constraints or fail to reach deep execution paths. We propose leveraging an N-gram model to learn SDRs from both kernel execution history and ongoing fuzzing results. This enables the fuzzer to capture dependencies in similar kernel versions while adapting to target-specific behaviors, thereby improving the validity of generated seeds. Additionally, we introduce a bidirectional Random Walk strategy to enhance the diversity of generated seeds. We implement this approach in a prototype, Psyzkaller, on top of Syzkaller. Experiments show that, trained with the large-scale DongTing dataset and continuously updated with ongoing fuzzing results, Psyzkaller improves Syzkaller's code coverage by 4.6%-7.0%, triggers 110.4%-187.2% more crashes, and discovers eight previously unknown kernel vulnerabilities. Furthermore, Psyzkaller outperforms state-of-the-art fuzzers such as ACTOR and SyzDescribe in both coverage and crashes.

cs.CR↗

arXiv2026

A First-Principles Derivation of LLM Policy Optimization: From Expected Reward to GRPO and Its Structural Extensions

Policy gradient algorithms for language models optimize the same objective $J(θ) = \mathbb{E}*{τ\sim p*θ(τ)}[R(τ)]$, which has exactly two factors: the trajectory probability $p_θ(τ)$ and the reward $R(τ)$. Every method from REINFORCE to PPO to GRPO and their descendants modifies one or both factors to address a specific failure in the preceding formulation. Existing surveys organize these methods by domain or chronology, which obscures the rationale behind each design choice and the precise location of its intervention within the gradient estimator. This survey revisits the landscape of LLM policy optimization from $J(θ)$ on first principles and uses the trajectory side, induced by $p_θ(τ)$, and the reward side, induced by $R(τ)$, as the two axes along which methods are located. It covers the path from REINFORCE and PPO to GRPO, as well as post-GRPO variants, Agentic RL, and GRPO-OPD. The resulting framework is unified, diagnostic, and extensible: it analyzes methods from a shared objective, identifies which side each method modifies and why, and applies the same trajectory and reward axes across these settings. Across these settings, the framework also exposes compound failures that no single-side fix resolves and that therefore require joint design of the trajectory side and the reward side. The boundary cases and coupled failures identified by this map mark where existing solutions run out and provide a principled starting point for designing the next generation of LLM policy optimization algorithms.

cs.AI↗

arXiv2026

DriveReward: A Comprehensive Dataset and Generative Vision-Language Reward Model for Autonomous Driving

Reward models play a pivotal role in reinforcement learning (RL) and multi-modal trajectory selection for autonomous driving. However, acquiring such rewards typically relies on hand-crafted rule-based objectives or perception ground truth, which hinders generalization for data-scaling. While Vision-Language Models (VLMs) have demonstrated feasibility as reward models in other domains, their effectiveness in driving tasks remains underexplored. In this work, we bridge this gap by (1) introducing DriveReward, a reasoning trajectory evaluation dataset rigorously labeled via temporally-grounded visual guidance, and augmented with counterfactual driving behaviors., (2) alongside a specialized Vision-Language Reward Model. To address the scarcity of failure cases in conventional datasets, we propose a counterfactual data annotation scheme to construct cases encompassing diverse driving styles and erroneous behaviors. Evaluations on our proposed benchmark reveal that even leading open-source and proprietary VLMs fail to excel across all tasks, highlighting significant room for improvement in existing models. Building on these findings, we subsequently tailor a specialized 1B reward model that outperforms larger VLMs on task-specific reward alignment. Finally, we validate our reward model's effectiveness by integrating it into RL finetuning and multi-modal trajectory scoring across multiple baselines, achieving performance comparable to rule-based reward calculations in both open-loop and closed-loop evaluation.

cs.CV↗

arXiv2026

PhoneHarness: Harnessing Phone-Use Agents through Mixed GUI, CLI, and Tool Actions

Phone agents are increasingly expected to complete real mobile workflows rather than merely predict the next screen action. However, much of the current mobile-agent literature still evaluates agents primarily as GUI controllers that observe a screen, emit taps and swipes, and are scored by target app state. Real phone-use tasks are broader: they require deciding when to use app GUIs, device-side commands, or structured tools, while leaving evidence that the intended side effect actually occurred. We introduce PhoneHarness, a mixed-action benchmark and execution harness for studying phone-use agents on verifiable mobile workflows. PhoneHarness runs a device-side agent loop over GUI, CLI, and host-side tool actions, combining deterministic action routing with bounded GUI delegation and auditable execution traces. Its benchmark, PhoneHarness Bench, evaluates whether agents complete tasks with observable side effects, not only whether they produce plausible final answers. On the annotated evaluation split, PhoneHarness reaches a 75.0% pass rate, outperforming the strongest non-PhoneHarness settings by 12.9 percentage points. PhoneHarness and PhoneHarness Bench therefore play distinct but mutually dependent roles: the harness makes mixed phone workflows executable, while the benchmark measures whether agents can use that harness reliably and safely. Our findings suggest that reliable phone automation depends on action-surface routing and verifiable execution, not only visual GUI control.

cs.CL↗

arXiv2026

Breaking Entropy Bounds: Accelerating RL Training via MTP with Rejection Sampling

Reinforcement learning (RL) has become a key component in modern large language models, yet the rollout stage remains the key bottleneck in RL training pipelines. Although Multi-Token Prediction (MTP) offers a natural solution to accelerate rollouts through speculative decoding, many studies have observed that MTP acceptance rates degrade significantly during RL training, leading to limited speedup performance. To address this bottleneck, we present Bebop, a systematic study of MTP in LLM post-training, and offer practical recipes to integrate MTP into large-scale RL pipelines. First, we reveal that the MTP acceptance rate is fundamentally bounded by the fluctuation of model entropy, which demonstrates a clear negative linear relationship with the rise of entropy in the RL stage. Second, we show that probabilistic rejection sampling largely alleviates the disturbance introduced by entropy in RL compared to greedy draft sampling. We further identify that the conventional MTP training objectives (cross-entropy or KL) are suboptimal in such settings, and therefore we propose a novel end-to-end TV loss that directly optimizes multi-step rejection sampling acceptance rate, yielding ~10% acceptance rate improvements, achieving up to 95% acceptance rates and up to 25% extra inference throughput gains across mathematical reasoning, code generation, and agentic tasks. Third, we test various online MTP training strategies during RL and show that pre-RL MTP training with e2e TV loss and rejection sampling achieves a consistent acceptance rate and speedup throughout the entire RL, eliminating the need for costly online MTP updating. We provide extensive experiments and analysis that validate our findings. Experimental results show our method achieves up to 1.8x end-to-end acceleration in async RL training of Qwen3.5, Qwen3.6, and Qwen3.7 models.

cs.LG↗

arXiv2026

Active-Passive Federated Learning for Vertically Partitioned Multi-view Data

Vertical federated learning is a natural and elegant approach to integrate multi-view data vertically partitioned across devices (clients) while preserving their privacies. Apart from the model training, existing methods requires the collaboration of all clients in the model inference. However, the model inference is probably maintained for service in a long time, while the collaboration, especially when the clients belong to different organizations, is unpredictable in real-world scenarios, such as concellation of contract, network unavailablity, etc., resulting in the failure of them. To address this issue, we, at the first attempt, propose a flexible Active-Passive Federated learning (APFed) framework. Specifically, the active client is the initiator of a learning task and responsible to build the complete model, while the passive clients only serve as assistants. Once the model built, the active client can make inference independently. In addition, we instance the APFed framework into two classification methods with employing the reconstruction loss and the contrastive loss on passive clients, respectively. Meanwhile, the two methods are tested in a set of experiments and achieves desired results, validating their effectiveness.

cs.LG↗

arXiv2026

ChinaHeritaQA: A Culturally-Grounded Visual Question Answering Dataset for World Heritage Sites in China

We introduce ChinaHeritaQA, a multimodal benchmark dataset for evaluating the cultural reasoning abilities of vision-language models (VLMs) on UNESCO World Heritage sites in China. The dataset comprises 2,279 in-the-wild images paired with 14,133 bilingual (Chinese/English) multiple-choice QA pairs spanning seven cognitive dimensions, from basic identity recognition to historical periodization and architectural analysis. Guided by a UNESCO-aligned heritage ontology and verified through rigorous human annotation, the dataset ensures linguistic quality and factual consistency. Evaluations of state-of-the-art VLMs reveal that while top models exceed human performance on average, substantial task-level variation emerges: models excel at visual recognition but struggle with culturally grounded reasoning. Performance also varies by dynasty and region. ChinaHeritaQA reveals that strong visual retrieval does not extend to cultural and historical understanding. We release the dataset to support future research on culturally aware multimodal learning.

cs.CV↗

arXiv2026

A decoupled energy-stable mixed finite element method for Poisson-Nernst-Planck-Navier-Stokes equations

We propose a novel linearized mixed finite element method for the Poisson-Nernst-Planck-Navier-Stokes (PNPNS) system. Specifically, the method combines a staggered time discretization that eliminates the need for expensive nonlinear solvers by carefully treating nonlinear terms in a time-staggered manner, with a mimetic spatial discretization that preserves the exact structure of the discrete de Rham complex. Both semi-discrete scheme and its fully discrete counterpart are developed, which preserve key physical properties, including conservation of the total mass and energy stability. Under appropriate assumptions on the initial data, a rigorous theoretical analysis is carried out for the fully discrete scheme. Numerical experiments using mimetic spectral elements are presented to demonstrate the properties and to verify the accuracy and effectiveness of the proposed decoupled approach.

math.NA↗

arXiv2026

FutureWeaver: Planning Test-Time Compute for Multi-Agent Systems with Modularized Collaboration

Scaling test-time computation has been shown to significantly improve large language model (LLM) performance without additional training. However, extending these techniques to multi-agent systems remains challenging: existing approaches lack principled mechanisms for allocating compute to enable effective collaboration, scaling coordination itself, or optimizing compute usage under explicit budget constraints. To address this gap, we propose FutureWeaver, a framework for planning and optimizing test-time compute allocation in multi-agent systems under fixed budgets. It introduces collaboration modules, formalized as modular, callable functions that encapsulate reusable multi-agent workflows and are automatically induced via self-play reflection from recurring interaction patterns. Building on these modules, it employs \emph{a dual-level planning architecture} that jointly performs short-horizon action selection and long-horizon abstract lookahead to optimize inference trajectories under budget constraints. Experiments on complex agent benchmarks demonstrate that FutureWeaver consistently outperforms baselines across diverse budget settings, validating its effectiveness for multi-agent collaboration in inference-time optimization.

cs.AI↗