arXiv Science⌕ Search

arXiv subjects

Yuxin Wang

Publications and source records attributed to Yuxin Wang.

At least 109 records · Page 6Linked to original sources
arXiv2025

GRIP: A Graph-Based Reasoning Instruction Producer

Large-scale, high-quality data is essential for advancing the reasoning capabilities of large language models (LLMs). As publicly available Internet data becomes increasingly scarce, synthetic data has emerged as a crucial research direction. However, existing data synthesis methods often suffer from limited scalability, insufficient sample diversity, and a tendency to overfit to seed data, which constrains their practical utility. In this paper, we present \textit{\textbf{GRIP}}, a \textbf{G}raph-based \textbf{R}easoning \textbf{I}nstruction \textbf{P}roducer that efficiently synthesizes high-quality and diverse reasoning instructions. \textit{GRIP} constructs a knowledge graph by extracting high-level concepts from seed data, and uniquely leverages both explicit and implicit relationships within the graph to drive large-scale and diverse instruction data synthesis, while employing open-source multi-model supervision to ensure data quality. We apply \textit{GRIP} to the critical and challenging domain of mathematical reasoning. Starting from a seed set of 7.5K math reasoning samples, we construct \textbf{GRIP-MATH}, a dataset containing 2.1 million synthesized question-answer pairs. Compared to similar synthetic data methods, \textit{GRIP} achieves greater scalability and diversity while also significantly reducing costs. On mathematical reasoning benchmarks, models trained with GRIP-MATH demonstrate substantial improvements over their base models and significantly outperform previous data synthesis methods.

cs.CL↗
arXiv2025

Recent progress in nickelate superconductors

The discovery of superconductivity in nickelate compounds has opened new avenues in the study of high-temperature superconductors. Here we provide a comprehensive overview of recent progress in the field, including all different nickelate systems, reduced-Ruddlesden-Popper-type infinite layer LaNiO$_2$, Ruddlesden-Popper-type bilayer La$_3$Ni$_2$O$_7$ and trilayer La$_4$Ni$_3$O$_{10}$. We begin by introducing the superconducting properties of the hole-doped LaNiO$_2$ system, which marked the starting point for nickelate superconductivity. We then turn to the bilayer La$_3$Ni$_2$O$_7$ system, discussing both its high-pressure and thin-film superconducting phases. This is followed by an examination of the trilayer La$_4$Ni$_3$O$_{10}$ system and other related multilayer nickelates. Throughout the review, we highlight emerging trends, key challenges, and open questions. We conclude by addressing current limitations in materials synthesis and characterization, and future directions that may help uncover the mechanisms driving superconductivity in these complex oxide systems.

cond-mat.supr-con↗
arXiv2025

VehicleWorld: A Highly Integrated Multi-Device Environment for Intelligent Vehicle Interaction

Intelligent vehicle cockpits present unique challenges for API Agents, requiring coordination across tightly-coupled subsystems that exceed typical task environments' complexity. Traditional Function Calling (FC) approaches operate statelessly, requiring multiple exploratory calls to build environmental awareness before execution, leading to inefficiency and limited error recovery. We introduce VehicleWorld, the first comprehensive environment for the automotive domain, featuring 30 modules, 250 APIs, and 680 properties with fully executable implementations that provide real-time state information during agent execution. This environment enables precise evaluation of vehicle agent behaviors across diverse, challenging scenarios. Through systematic analysis, we discovered that direct state prediction outperforms function calling for environmental control. Building on this insight, we propose State-based Function Call (SFC), a novel approach that maintains explicit system state awareness and implements direct state transitions to achieve target conditions. Experimental results demonstrate that SFC significantly outperforms traditional FC approaches, achieving superior execution accuracy and reduced latency. We have made all implementation code publicly available on Github https://github.com/OpenMOSS/VehicleWorld.

cs.AI↗
arXiv2025

Novel constants based on the generalization of Von Neumann-Jordan constant

We introduce a new geometric constant based on a generalization of the parallelogram law, and study its properties as well as some relationships with other well-known geometric constants. A sufficient condition for normal structure is presented. Next, we introduce a constant and calculate its value in a specific space. Furthermore, we introduce another new constant and investigate some of its basic properties.

math.FA↗
arXiv2025

Skew von Neumann constant in Weak Orlicz Spaces and Weak Lebesgue Spaces

This paper defines the skew von Neumann constant in quasi-Banach spaces. Meanwhile, we obtain two constants. It presents the upper and lower bounds of two constants. Subsequently, it deduces the lower bound of the skew von Neumann constant within weak Orlicz spaces. Then, leveraging the relationship between weak Orlicz spaces and weak Lebesgue spaces, the lower bound of the skew von Neumann - Jordan constant in weak Lebesgue spaces is established. Meanwhile, in this paper, we also generalize the skew von Neumann constant. Then we present the lower bounds for the $p$-th von Neumann constant in weak Orlicz spaces and weak Lebesgue spaces.

math.FA↗
arXiv2025

IGD: Instructional Graphic Design with Multimodal Layer Generation

Graphic design visually conveys information and data by creating and combining text, images and graphics. Two-stage methods that rely primarily on layout generation lack creativity and intelligence, making graphic design still labor-intensive. Existing diffusion-based methods generate non-editable graphic design files at image level with poor legibility in visual text rendering, which prevents them from achieving satisfactory and practical automated graphic design. In this paper, we propose Instructional Graphic Designer (IGD) to swiftly generate multimodal layers with editable flexibility with only natural language instructions. IGD adopts a new paradigm that leverages parametric rendering and image asset generation. First, we develop a design platform and establish a standardized format for multi-scenario design files, thus laying the foundation for scaling up data. Second, IGD utilizes the multimodal understanding and reasoning capabilities of MLLM to accomplish attribute prediction, sequencing and layout of layers. It also employs a diffusion model to generate image content for assets. By enabling end-to-end training, IGD architecturally supports scalability and extensibility in complex graphic design tasks. The superior experimental results demonstrate that IGD offers a new solution for graphic design.

cs.CV↗
arXiv2025

Failed superconductivity in a Mott spin liquid material

A central challenge for understanding unconventional superconductivity in most strongly correlated electronic materials is their complicated band structure and presence of competing orders. In contrast, quasi-two-dimensional organic spin liquids are single-band systems with superconductivity arising near the bandwidth-tuned Mott metal-insulator transition in the absence of other orders. Here, we study chemically substituted $κ$-organics in which superconducting fluctuations emerge in the phase coexistence region between the Mott insulator and the Fermi liquid. Using magnetotransport and ac susceptibility measurements, we find that global superconductivity fails to set in as temperature $T\rightarrow 0$. Our results indicate instead the presence of superconducting domains embedded in the metallic percolating cluster that undergo a magnetic field-tuned quantum superconductor-to-metal phase transition. Surprisingly, albeit consistent with the percolation picture, universal conductance fluctuations are seen at high fields in macroscopic samples. The observed interplay of the intrinsic inhomogeneity and quantum phase fluctuations provides a new insight into failed superconductivity, a phenomenon seen in various conventional and unconventional superconductors, including cuprates.

cond-mat.supr-con↗
arXiv2025

Test-Time Scaling with Reflective Generative Model

We introduce our first reflective generative model MetaStone-S1, which obtains OpenAI o3-mini's performance via the new Reflective Generative Form. The new form focuses on high-quality reasoning trajectory selection and contains two novelties: 1) A unified interface for policy and process reward model: we share the backbone network and use task-specific heads for reasoning trajectory predicting and scoring respectively, introducing only 53M extra parameters for trajectory scoring. 2) Eliminating the reliance on process-level annotation: we provide a self-supervised process reward model, which can directly learn the high-quality reasoning trajectory selection from the outcome reward. Equipped with the reflective generative form, MetaStone-S1 is naturally suitable for test-time scaling, and we provide three reasoning effort modes (low, medium, and high) based on the controllable thinking length. Experiments demonstrate that our MetaStone-S1 achieves comparable performance to OpenAI o3-mini's series with only 32B parameter size. To support the research community, we have open-sourced MetaStone-S1 at https://github.com/MetaStone-AI/MetaStone-S1.

cs.LG↗
arXiv2025

On the equivalent p-th von Neumann-Jordan constant associated with isosceles orthogonality in Banach spaces

In this paper, we define a new geometric constant based on isosceles orthogonality, denoted by . Through research, we find that this constant is the equivalent p-th von Neumann Jordan constant in the sense of isosceles orthogonality. First, we obtain some basic properties of the constant. Then, we calculate the upper and lower bounds of the constant. Through three examples, it is found that the upper bound of the constant is attainable. We also compare the relationship between this constant and other constants. Finally, we establish the connection between the constant and some geometric properties in Banach spaces, such as uniform non-squareness, uniform smoothness.

math.FA↗
arXiv2025

BurstGPT: A Real-world Workload Dataset to Optimize LLM Serving Systems

Serving systems for Large Language Models (LLMs) are often optimized to improve quality of service (QoS) and throughput. However, due to the lack of open-source LLM serving workloads, these systems are frequently evaluated under unrealistic workload assumptions. Consequently, performance may degrade when systems are deployed in real-world scenarios. This work presents BurstGPT, an LLM serving workload with 10.31 million traces from regional Azure OpenAI GPT services over 213 days. BurstGPT captures LLM serving characteristics from user, model and system perspectives: (1) User request concurrency: burstiness variations of requests in Azure OpenAI GPT services, revealing diversified concurrency patterns in different services and model types. (2) User conversation patterns: counts and intervals within conversations for service optimizations. (3) Model response lengths: auto-regressive serving processes of GPT models, showing statistical relations between requests and their responses. (4) System response failures: failures of conversation and API services, showing intensive resource needs and limited availability of LLM services in Azure. The details of the characteristics can serve multiple purposes in LLM serving optimizations, such as system evaluation and trace provisioning. In our demo evaluation with BurstGPT, frequent variations in BurstGPT reveal declines in efficiency, stability, or reliability in realistic LLM serving. We identify that the generalization of KV cache management, scheduling and disaggregation optimizations can be improved under realistic workload evaluations. BurstGPT is publicly available now at https://github.com/HPMLL/BurstGPT and is widely used to develop prototypes of LLM serving frameworks in the industry.

cs.DC↗
arXiv2025

FamilyTool: A Multi-hop Personalized Tool Use Benchmark

The integration of tool learning with Large Language Models (LLMs) has expanded their capabilities in handling complex tasks by leveraging external tools. However, existing benchmarks for tool learning inadequately address critical real-world personalized scenarios, particularly those requiring multi-hop reasoning and inductive knowledge adaptation in dynamic environments. To bridge this gap, we introduce FamilyTool, a novel benchmark grounded in a family-based knowledge graph (KG) that simulates personalized, multi-hop tool use scenarios. FamilyTool, including base and extended datasets, challenges LLMs with queries spanning from 1 to 4 relational hops (e.g., inferring familial connections and preferences) and 2 to 6 hops respectively, and incorporates an inductive KG setting where models must adapt to unseen user preferences and relationships without re-training, a common limitation in prior approaches that compromises generalization. We further propose KGETool: a simple KG-augmented evaluation pipeline to systematically assess LLMs' tool use ability in these settings. Experiments reveal significant performance gaps in state-of-the-art LLMs, with accuracy dropping sharply as hop complexity increases and inductive scenarios exposing severe generalization deficits. These findings underscore the limitations of current LLMs in handling personalized, evolving real-world contexts and highlight the urgent need for advancements in tool-learning frameworks. FamilyTool serves as a critical resource for evaluating and advancing LLM agents' reasoning, adaptability, and scalability in complex, dynamic environments. Code and dataset are available at \href{https://github.com/yxzwang/FamilyTool}{https://github.com/yxzwang/FamilyTool}.

cs.AI↗
arXiv2025

On some generalized geometric constants with two parameters in Banach spaces

In this paper, we build upon the TX constant that was introduced by Alonso and Llorens-Fuster in 2008. Through the incorporation of suitable parameters, we have successfully generalized the aforementioned constant into two novel forms of geometric constants, which are denoted as T1(λ,μ,X ) and T2(\k{appa},τ,X ). First, we obtained some basic properties of these two constants, such as the upper and lower bounds. Next, these two constants served as the basis for our characterization of Hilbert spaces. More significantly, our findings reveal that these two constants exhibit a profound and intricate interrelation with other well-known constants in Banach spaces. Finally, we characterized uniformly non-square spaces by means of these two constants.

math.FA↗
arXiv2025

The skew generalized von Neumann-Jordan type constant in Banach spaces

Recently, the von Neumann-Jordan type constants C(X) has defined by Takahashi. A new skew generalized constant Cp(λ,μ,X) based on C(X) constant is given in this paper. First, we will obtain some basic properties of this new constant. Moreover, some relations between this new constant and other constants are investigated. Specially, with the Banach-Mazur distance, we use this new constant to study isomorphic Banach spaces. Ultimately, by leveraging the connection between the newly introduced constant and the weak orthogonality coefficient ω(X), a sufficient condition for normal structure is established.

math.FA↗
arXiv2025

Inference of high-dimensional weak instrumental variable regression models without ridge-regularization

Inference of instrumental variable regression models with many weak instruments attracts many attentions recently. To extend the classical Anderson-Rubin test to high-dimensional setting, many procedures adopt ridge-regularization. However, we show that it is not necessary to consider ridge-regularization. Actually we propose a new quadratic-type test statistic which does not involve tuning parameters. Our quadratic-type test exhibits high power against dense alternatives. While for sparse alternatives, we derive the asymptotic distribution of an existing maximum-type test, enabling the use of less conservative critical values. To achieve strong performance across a wide range of scenarios, we further introduce a combined test procedure that integrates the strengths of both approaches. This combined procedure is powerful without requiring prior knowledge of the underlying sparsity of the first-stage model. Compared to existing methods, our proposed tests are easy to implement, free of tuning parameters, and robust to arbitrarily weak instruments as well as heteroskedastic errors. Simulation studies and empirical applications demonstrate the advantages of our methods over existing approaches.

stat.ME↗
arXiv2025

The electronic structure and disorder effect of La$_3$Ni$_2$O$_{7}$ superconductor

Determining the electronic structure of La$_3$Ni$_2$O$_7$ is an essential step towards uncovering their superconducting mechanism. It is widely believed that the bilayer apical oxygens play an important role in the bilayer La$_3$Ni$_2$O$_7$ electronic structure. Applying the hybrid exchange-correlation functionals, we obtain a more accurate electronic structure of La$_3$Ni$_2$O$_7$ at its high-pressure phase, where the binding $d_{z^2}$ band is below the Fermi level owing to apical oxygen. The symmetry properties of this electronic structure and its corresponding tight-binding model are further analyzed. We find the antisymmetric part is highly entangled leading to a minimal nearly degenerate two-orbital model. Then, the apical oxygen vacancies effect is studied using the dynamical cluster approximation. This disorder effect strongly destroys the antisymmetric $β$ Fermi surface leading to the possible disappearance of superconductivity.

cond-mat.supr-con↗
arXiv2025

Some aspects of generalized Dunkl-Williams constant in Banach spaces

This article delves into an exploration of two innovative constants, namely DW(X,α,\b{eta}) and DWB (X,α,\b{eta}), both of which constitute extensions of the Dunkl-Williams constant. We derive both the upper and lower bounds for these two constants and establish two equivalent relations between them. Moreover, we elucidate the relationships between these constants and several well-known constants. Additionally, we have refined the value of the DWB (X,α,\b{eta}) constant in certain specific Banach spaces.

math.FA↗
arXiv2025

A new geometric constant to compare p-angular and skew p-angular distances

The $p$-angular distance was first introduced by Maligranda in 2006, while the skew $p$-angular distance was first introduced by Rooin in 2018. In this paper, we shall introduce a new geometric constant named Maligranda-Rooin constant in Banach spaces to compare $p$-angular distance and skew $p$-angular distance. We denote the Maligranda-Rooin constant as $\mathcal{M} \mathcal{R}_p(\mathcal{X})$. First, the upper and lower bounds for the $\mathcal{M} \mathcal{R}_p(\mathcal{X})$ constant is given. Next, it's shown that, a normed linear space is an inner space if and only if $\mathcal{M} \mathcal{R}_p(\mathcal{X})=1$. Moreover, an equivalent form of this new constant is established. By means of the $\mathcal{M} \mathcal{R}_p(\mathcal{X})$ constant, we carry out the quantification of the characterization of uniform nonsquareness. Finally, we study the relationship between the $\mathcal{M} \mathcal{R}_p(\mathcal{X})$ constant, uniform convexity, uniform smooth and normal structure.

math.FA↗