arXiv subjects

Xinyu Wang

Publications and source records attributed to Xinyu Wang.

At least 217 records · Page 12Linked to original sources

arXiv2025

Graph Neural Network Enhanced Sequential Recommendation Method for Cross-Platform Ad Campaign

In order to improve the accuracy of cross-platform advertisement recommendation, a graph neural network (GNN)- based advertisement recommendation method is analyzed. Through multi-dimensional modeling, user behavior data (e.g., click frequency, active duration) reveal temporal patterns of interest evolution, ad content (e.g., type, tag, duration) influences semantic preferences, and platform features (e.g., device type, usage context) shape the environment where interest transitions occur. These factors jointly enable the GNN to capture the latent pathways of user interest migration across platforms. The experimental results are based on the datasets of three platforms, and Platform B reaches 0.937 in AUC value, which is the best performance. Platform A and Platform C showed a slight decrease in precision and recall with uneven distribution of ad labels. By adjusting the hyperparameters such as learning rate, batch size and embedding dimension, the adaptability and robustness of the model in heterogeneous data are further improved.

cs.LG↗

arXiv2025

Emergent dynamics of spatially extended relativistic kinetic Cucker-Smale model

We study the emergent dynamics of the relativistic kinetic Cucker-Smale (RKCS) model without assuming compactness in spatial and velocity support. In this setting, the lower bound of the kernel function in the nonlocal velocity alignment force can be zero so that the previous approach based on the energy method does not provide a quantitative flocking estimate. To overcome this difficulty, we introduce a suitable decay ansatz for the one-particle distribution function and an {\it effective domain} by identifying a time-varying region in which the total mass outside of it decays to zero asymptotically. Using these two ingredients, we show that weak flocking dynamics emerges asymptotically in the sense that the second moment for the velocity fluctuation around the velocity average tends to zero asymptotically, whereas the second moment for spatial fluctuations around the center of mass remains bounded uniformly in time. Our results demonstrate the robustness of the emergent dynamics in the RKCS model across various non-compact physically important distributions, including Gaussian, sub-Gaussian, and $D$-th moment integrable distributions.

math.AP↗

arXiv2025

WebSailor: Navigating Super-human Reasoning for Web Agent

Transcending human cognitive limitations represents a critical frontier in LLM training. Proprietary agentic systems like DeepResearch have demonstrated superhuman capabilities on extremely complex information-seeking benchmarks such as BrowseComp, a feat previously unattainable. We posit that their success hinges on a sophisticated reasoning pattern absent in open-source models: the ability to systematically reduce extreme uncertainty when navigating vast information landscapes. Based on this insight, we introduce WebSailor, a complete post-training methodology designed to instill this crucial capability. Our approach involves generating novel, high-uncertainty tasks through structured sampling and information obfuscation, RFT cold start, and an efficient agentic RL training algorithm, Duplicating Sampling Policy Optimization (DUPO). With this integrated pipeline, WebSailor significantly outperforms all opensource agents in complex information-seeking tasks, matching proprietary agents' performance and closing the capability gap.

cs.CL↗

arXiv2025

TinyAlign: Boosting Lightweight Vision-Language Models by Mitigating Modal Alignment Bottlenecks

Lightweight Vision-Language Models (VLMs) are indispensable for resource-constrained applications. The prevailing approach to aligning vision and language models involves freezing both the vision encoder and the language model while training small connector modules. However, this strategy heavily depends on the intrinsic capabilities of the language model, which can be suboptimal for lightweight models with limited representational capacity. In this work, we investigate this alignment bottleneck through the lens of mutual information, demonstrating that the constrained capacity of the language model inherently limits the Effective Mutual Information (EMI) between multimodal inputs and outputs, thereby compromising alignment quality. To address this challenge, we propose TinyAlign, a novel framework inspired by Retrieval-Augmented Generation, which strategically retrieves relevant context from a memory bank to enrich multimodal inputs and enhance their alignment. Extensive empirical evaluations reveal that TinyAlign significantly reduces training loss, accelerates convergence, and enhances task performance. Remarkably, it allows models to achieve baseline-level performance with only 40\% of the fine-tuning data, highlighting exceptional data efficiency. Our work thus offers a practical pathway for developing more capable lightweight VLMs while introducing a fresh theoretical lens to better understand and address alignment bottlenecks in constrained multimodal systems.

cs.LG↗

arXiv2025

Social Scientists on the Role of AI in Research

The integration of artificial intelligence (AI) into social science research practices raises significant technological, methodological, and ethical issues. We present a community-centric study drawing on 284 survey responses and 15 semi-structured interviews with social scientists, describing their familiarity with, perceptions of the usefulness of, and ethical concerns about the use of AI in their field. A crucial innovation in study design is to split our survey sample in half, providing the same questions to each -- but randomizing whether participants were asked about "AI" or "Machine Learning" (ML). We find that the use of AI in research settings has increased significantly among social scientists in step with the widespread popularity of generative AI (genAI). These tools have been used for a range of tasks, from summarizing literature reviews to drafting research papers. Some respondents used these tools out of curiosity but were dissatisfied with the results, while others have now integrated them into their typical workflows. Participants, however, also reported concerns with the use of AI in research contexts. This is a departure from more traditional ML algorithms which they view as statistically grounded. Participants express greater trust in ML, citing its relative transparency compared to black-box genAI systems. Ethical concerns, particularly around automation bias, deskilling, research misconduct, complex interpretability, and representational harm, are raised in relation to genAI. To guide this transition, we offer recommendations for AI developers, researchers, educators, and policymakers focusing on explainability, transparency, ethical safeguards, sustainability, and the integration of lived experiences into AI design and evaluation processes.

cs.CY↗

arXiv2025

Enumerating several statistics of r-Colored Dyck paths with no dd-steps having the same colors

An $r$-colored Dyck path is a Dyck path with all $\mathbf{d}$-steps having one of $r$ colors in $[r]=\{1, 2, \dots, r\}$. In this paper, we consider several statistics on the set $\mathcal{A}_{n,0}^{(r)}$ of $r$-colored Dyck paths of length $2n$ with no two consecutive $\mathbf{d}$-steps having the same colors. Precisely, the paper studies the statistics ``number of points" at level $\ell$, ``number of $\mathbf{u}$-steps" at level $\ell+1$, ``number of peaks" at level $\ell+1$ and ``number of $\mathbf{udu}$-steps" on the set $\mathcal{A}_{n,0}^{(r)}$. The counting formulas of the first three statistics are established by Riordan arrays related to $S(a,b; x)$, the weighted generating function of $(a,b)$-Schröder paths. By a useful and surprising relations satisfied by $S(a,b; x)$, several identities related to these counting formulas are also described.

math.CO↗

arXiv2025

PECAN: LLM-Guided Dynamic Progress Control with Attention-Guided Hierarchical Weighted Graph for Long-Document QA

Long-document QA presents challenges with large-scale text and long-distance dependencies. Recent advances in Large Language Models (LLMs) enable entire documents to be processed in a single pass. However, their computational cost is significantly high. Retrieval-Augmented Generation (RAG) methods split text into smaller chunks, but they often yield inferior results and may lose global context. Recent approaches that integrate LLMs into RAG via iterative summarization either underutilize LLM capabilities or still incur high computational costs. In this paper, we combine the high accuracy of LLMs with the efficiency of RAG and propose LLM-Guided Dynamic Progress Control with Attention-Based Hierarchical Weighted Graph (PECAN). Our method introduces two key improvements: (1) LLM-Guided Dynamic Progress Control: We leverage LLMs to dynamically control the retrieval process, adjusting the amount of retrieved information based on different queries to achieve a better balance of effectiveness and efficiency. (2) Attention-Guided Retrieval: We propose a novel retrieval method that constructs a hierarchical graph where edges are derived by LLM attention weights. Experimental results demonstrate that PECAN achieves LLM-level performance while maintaining computational complexity comparable to that of RAG methods on two single-document and two multi-document QA datasets.

cs.CL↗

arXiv2025

OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning

Scoring the Optical Character Recognition (OCR) capabilities of Large Multimodal Models (LMMs) has witnessed growing interest. Existing benchmarks have highlighted the impressive performance of LMMs in text recognition; however, their abilities in certain challenging tasks, such as text localization, handwritten content extraction, and logical reasoning, remain underexplored. To bridge this gap, we introduce OCRBench v2, a large-scale bilingual text-centric benchmark with currently the most comprehensive set of tasks (4x more tasks than the previous multi-scene benchmark OCRBench), the widest coverage of scenarios (31 diverse scenarios), and thorough evaluation metrics, with 10,000 human-verified question-answering pairs and a high proportion of difficult samples. Moreover, we construct a private test set with 1,500 manually annotated images. The consistent evaluation trends observed across both public and private test sets validate the OCRBench v2's reliability. After carefully benchmarking state-of-the-art LMMs, we find that most LMMs score below 50 (100 in total) and suffer from five-type limitations, including less frequently encountered text recognition, fine-grained perception, layout perception, complex element parsing, and logical reasoning. The project website is at: https://99franklin.github.io/ocrbench_v2/

cs.CV↗

arXiv2025

NavBench: Probing Multimodal Large Language Models for Embodied Navigation

Multimodal Large Language Models (MLLMs) have demonstrated strong generalization in vision-language tasks, yet their ability to understand and act within embodied environments remains underexplored. We present NavBench, a benchmark to evaluate the embodied navigation capabilities of MLLMs under zero-shot settings. NavBench consists of two components: (1) navigation comprehension, assessed through three cognitively grounded tasks including global instruction alignment, temporal progress estimation, and local observation-action reasoning, covering 3,200 question-answer pairs; and (2) step-by-step execution in 432 episodes across 72 indoor scenes, stratified by spatial, cognitive, and execution complexity. To support real-world deployment, we introduce a pipeline that converts MLLMs' outputs into robotic actions. We evaluate both proprietary and open-source models, finding that GPT-4o performs well across tasks, while lighter open-source models succeed in simpler cases. Results also show that models with higher comprehension scores tend to achieve better execution performance. Providing map-based context improves decision accuracy, especially in medium-difficulty scenarios. However, most models struggle with temporal understanding, particularly in estimating progress during navigation, which may pose a key challenge.

cs.CV↗

arXiv2025

Benchmarking Multimodal Retrieval Augmented Generation with Dynamic VQA Dataset and Self-adaptive Planning Agent

Multimodal Retrieval Augmented Generation (mRAG) plays an important role in mitigating the "hallucination" issue inherent in multimodal large language models (MLLMs). Although promising, existing heuristic mRAGs typically predefined fixed retrieval processes, which causes two issues: (1) Non-adaptive Retrieval Queries. (2) Overloaded Retrieval Queries. However, these flaws cannot be adequately reflected by current knowledge-seeking visual question answering (VQA) datasets, since the most required knowledge can be readily obtained with a standard two-step retrieval. To bridge the dataset gap, we first construct Dyn-VQA dataset, consisting of three types of "dynamic" questions, which require complex knowledge retrieval strategies variable in query, tool, and time: (1) Questions with rapidly changing answers. (2) Questions requiring multi-modal knowledge. (3) Multi-hop questions. Experiments on Dyn-VQA reveal that existing heuristic mRAGs struggle to provide sufficient and precisely relevant knowledge for dynamic questions due to their rigid retrieval processes. Hence, we further propose the first self-adaptive planning agent for multimodal retrieval, OmniSearch. The underlying idea is to emulate the human behavior in question solution which dynamically decomposes complex multimodal questions into sub-question chains with retrieval action. Extensive experiments prove the effectiveness of our OmniSearch, also provide direction for advancing mRAG. The code and dataset will be open-sourced at https://github.com/Alibaba-NLP/OmniSearch.

cs.CL↗

arXiv2025

Qwen3 Technical Report

In this work, we present Qwen3, the latest version of the Qwen model family. Qwen3 comprises a series of large language models (LLMs) designed to advance performance, efficiency, and multilingual capabilities. The Qwen3 series includes models of both dense and Mixture-of-Expert (MoE) architectures, with parameter scales ranging from 0.6 to 235 billion. A key innovation in Qwen3 is the integration of thinking mode (for complex, multi-step reasoning) and non-thinking mode (for rapid, context-driven responses) into a unified framework. This eliminates the need to switch between different models--such as chat-optimized models (e.g., GPT-4o) and dedicated reasoning models (e.g., QwQ-32B)--and enables dynamic mode switching based on user queries or chat templates. Meanwhile, Qwen3 introduces a thinking budget mechanism, allowing users to allocate computational resources adaptively during inference, thereby balancing latency and performance based on task complexity. Moreover, by leveraging the knowledge from the flagship models, we significantly reduce the computational resources required to build smaller-scale models, while ensuring their highly competitive performance. Empirical evaluations demonstrate that Qwen3 achieves state-of-the-art results across diverse benchmarks, including tasks in code generation, mathematical reasoning, agent tasks, etc., competitive against larger MoE models and proprietary models. Compared to its predecessor Qwen2.5, Qwen3 expands multilingual support from 29 to 119 languages and dialects, enhancing global accessibility through improved cross-lingual understanding and generation capabilities. To facilitate reproducibility and community-driven research and development, all Qwen3 models are publicly accessible under Apache 2.0.

cs.CL↗

arXiv2025

ActRef: Enhancing the Understanding of Python Code Refactoring with Action-Based Analysis

Refactoring, the process of improving the code structure of a software system without altering its behavior, is crucial for managing code evolution in software development. Identifying refactoring actions in source code is essential for understanding software evolution and guiding developers in maintaining and improving the code quality. This study presents an action-based Refactoring Analysis Framework named ActRef, a novel algorithm designed to advance the detection and understanding of Python refactorings through a unique code change action-based analysis of code changes. ActRef mining multiple refactoring types (e.g., move, rename, extract, and inline operations) based on diff actions, covering multiple granularity levels including variable, method, class, and module levels. By focusing on the code change actions, ActRef provides a Python-adaptive solution to detect intricate refactoring patterns. Our evaluation, conducted on 1,914 manually validated refactoring instances from 136 open-source Python projects. The evaluation results show that ActRef achieves high precision(0.80) and recall(0.92), effectively identifying multiple refactoring types. Compared with leading baselines, including PyRef, PyRef with MLRefScanner, DeepSeek-R1 and ChatGPT-4, ActRef consistently demonstrates superior performance in detecting Python refactorings across various types. While matching PyRef in runtime efficiency, ActRef supports a broader spectrum of refactoring types and more refactoring mining levels. ActRef shows an effective and scalable approach for mining refactorings in dynamic Python codebases and introduces a new perspective on understanding code.

cs.SE↗

arXiv2025

Leveraging Partial SMILES Validation Scheme for Enhanced Drug Design in Reinforcement Learning Frameworks

SMILES-based molecule generation has emerged as a powerful approach in drug discovery. Deep reinforcement learning (RL) using large language model (LLM) has been incorporated into the molecule generation process to achieve high matching score in term of likelihood of desired molecule candidates. However, a critical challenge in this approach is catastrophic forgetting during the RL phase, where knowledge such as molecule validity, which often exceeds 99\% during pretraining, significantly deteriorates. Current RL algorithms applied in drug discovery, such as REINVENT, use prior models as anchors to retian pretraining knowledge, but these methods lack robust exploration mechanisms. To address these issues, we propose Partial SMILES Validation-PPO (PSV-PPO), a novel RL algorithm that incorporates real-time partial SMILES validation to prevent catastrophic forgetting while encouraging exploration. Unlike traditional RL approaches that validate molecule structures only after generating entire sequences, PSV-PPO performs stepwise validation at each auto-regressive step, evaluating not only the selected token candidate but also all potential branches stemming from the prior partial sequence. This enables early detection of invalid partial SMILES across all potential paths. As a result, PSV-PPO maintains high validity rates even during aggressive exploration of the vast chemical space. Our experiments on the PMO and GuacaMol benchmark datasets demonstrate that PSV-PPO significantly reduces the number of invalid generated structures while maintaining competitive exploration and optimization performance. While our work primarily focuses on maintaining validity, the framework of PSV-PPO can be extended in future research to incorporate additional forms of valuable domain knowledge, further enhancing reinforcement learning applications in drug discovery.

cs.LG↗

arXiv2025

Algorithmic Detection of Jacobi Stability for Systems of Second Order Differential Equations

This paper introduces an algorithmic approach to the analysis of Jacobi stability of systems of second order ordinary differential equations (ODEs) via the Kosambi--Cartan--Chern (KCC) theory. We develop an efficient symbolic program using Maple for computing the second KCC invariant for systems of second order ODEs in arbitrary dimension. The program allows us to systematically analyze Jacobi stability of a system of second order ODEs by means of real solving and solution classification using symbolic computation. The effectiveness of the proposed approach is illustrated by a model of wound strings, a two-dimensional airfoil model with cubic nonlinearity in supersonic flow and a 3-DOF tractor seat-operator model. The computational results on Jacobi stability of these models are further verified by numerical simulations. Moreover, our algorithmic approach allows us to detect hand-guided computation errors in published papers.

cs.SC↗

arXiv2025

QuantBench: Benchmarking AI Methods for Quantitative Investment

The field of artificial intelligence (AI) in quantitative investment has seen significant advancements, yet it lacks a standardized benchmark aligned with industry practices. This gap hinders research progress and limits the practical application of academic innovations. We present QuantBench, an industrial-grade benchmark platform designed to address this critical need. QuantBench offers three key strengths: (1) standardization that aligns with quantitative investment industry practices, (2) flexibility to integrate various AI algorithms, and (3) full-pipeline coverage of the entire quantitative investment process. Our empirical studies using QuantBench reveal some critical research directions, including the need for continual learning to address distribution shifts, improved methods for modeling relational financial data, and more robust approaches to mitigate overfitting in low signal-to-noise environments. By providing a common ground for evaluation and fostering collaboration between researchers and practitioners, QuantBench aims to accelerate progress in AI for quantitative investment, similar to the impact of benchmark platforms in computer vision and natural language processing.

q-fin.CP↗

arXiv2025

Polygon: Symbolic Reasoning for SQL using Conflict-Driven Under-Approximation Search

We present a novel symbolic reasoning engine for SQL which can efficiently generate an input $I$ for $n$ queries $P_1, \cdots, P_n$, such that their outputs on $I$ satisfy a given property (expressed in SMT). This is useful in different contexts, such as disproving equivalence of two SQL queries and disambiguating a set of queries. Our first idea is to reason about an under-approximation of each $P_i$ -- that is, a subset of $P_i$'s input-output behaviors. While it makes our approach both semantics-aware and lightweight, this idea alone is incomplete (as a fixed under-approximation might miss some behaviors of interest). Therefore, our second idea is to perform search over an expressive family of under-approximations (which collectively cover all program behaviors of interest), thereby making our approach complete. We have implemented these ideas in a tool, Polygon, and evaluated it on over 30,000 benchmarks across two tasks (namely, SQL equivalence refutation and query disambiguation). Our evaluation results show that Polygon significantly outperforms all prior techniques.

cs.PL↗

arXiv2025

Bayesian method for quantifying the non-Gaussian fluctuations in low-intermediate energy heavy ion collisions

In this work, we present a model-independent method to quantify the non-Gaussian fluctuations in the observable distributions, which are assessed by the difference between the measured observable distributions and reconstructed observable distributions via the Bayesian method. Our results indicate that the strength of non-Gaussian fluctuation increases with the beam energy, and is primarily driven by non-central collision mechanisms. The experimental measurement of the strength of non-Gaussian fluctuation of the observable distributions will provide valuable insights into understanding the nonequilibrium effects in heavy ion collisions and the liquid-gas phase transition in finite nuclei.

nucl-th↗

arXiv2025

Monolingual and Multilingual Misinformation Detection for Low-Resource Languages: A Comprehensive Survey

In today's global digital landscape, misinformation transcends linguistic boundaries, posing a significant challenge for moderation systems. Most approaches to misinformation detection are monolingual, focused on high-resource languages, i.e., a handful of world languages that have benefited from substantial research investment. This survey provides a comprehensive overview of the current research on misinformation detection in low-resource languages, both in monolingual and multilingual settings. We review existing datasets, methodologies, and tools used in these domains, identifying key challenges related to: data resources, model development, cultural and linguistic context, and real-world applications. We examine emerging approaches, such as language-generalizable models and multi-modal techniques, and emphasize the need for improved data collection practices, interdisciplinary collaboration, and stronger incentives for socially responsible AI research. Our findings underscore the importance of systems capable of addressing misinformation across diverse linguistic and cultural contexts.

cs.CL↗