arXiv subjects

Hao Zhou

Publications and source records attributed to Hao Zhou.

At least 199 records · Page 11Linked to original sources

arXiv2025

DAPO: An Open-Source LLM Reinforcement Learning System at Scale

Inference scaling empowers LLMs with unprecedented reasoning ability, with reinforcement learning as the core technique to elicit complex reasoning. However, key technical details of state-of-the-art reasoning LLMs are concealed (such as in OpenAI o1 blog and DeepSeek R1 technical report), thus the community still struggles to reproduce their RL training results. We propose the $\textbf{D}$ecoupled Clip and $\textbf{D}$ynamic s$\textbf{A}$mpling $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{DAPO}$) algorithm, and fully open-source a state-of-the-art large-scale RL system that achieves 50 points on AIME 2024 using Qwen2.5-32B base model. Unlike previous works that withhold training details, we introduce four key techniques of our algorithm that make large-scale LLM RL a success. In addition, we open-source our training code, which is built on the verl framework, along with a carefully curated and processed dataset. These components of our open-source system enhance reproducibility and support future research in large-scale LLM RL.

cs.LG↗

arXiv2025

Understanding 6G through Language Models: A Case Study on LLM-aided Structured Entity Extraction in Telecom Domain

Knowledge understanding is a foundational part of envisioned 6G networks to advance network intelligence and AI-native network architectures. In this paradigm, information extraction plays a pivotal role in transforming fragmented telecom knowledge into well-structured formats, empowering diverse AI models to better understand network terminologies. This work proposes a novel language model-based information extraction technique, aiming to extract structured entities from the telecom context. The proposed telecom structured entity extraction (TeleSEE) technique applies a token-efficient representation method to predict entity types and attribute keys, aiming to save the number of output tokens and improve prediction accuracy. Meanwhile, TeleSEE involves a hierarchical parallel decoding method, improving the standard encoder-decoder architecture by integrating additional prompting and decoding strategies into entity extraction tasks. In addition, to better evaluate the performance of the proposed technique in the telecom domain, we further designed a dataset named 6GTech, including 2390 sentences and 23747 words from more than 100 6G-related technical publications. Finally, the experiment shows that the proposed TeleSEE method achieves higher accuracy than other baseline techniques, and also presents 5 to 9 times higher sample processing speed.

cs.CL↗

arXiv2025

Thermal boundary conductance in standalone and non-standalone GaN/AlN heterostructures predicted using machine learning interatomic potentials

GaN/AlN interfaces are essential in advanced high-power and high-frequency electronic devices, where effective thermal management is crucial for optimal performance and reliability. This work investigates the thermal boundary conductance (TBC) of standalone and non-standalone GaN/AlN heterostructures using non-equilibrium molecular dynamics (NEMD) driven by accurate machine learning interatomic potentials trained from density-functional theory calculations. For the standalone interface, the TBC is found to be ~600 MW m-2 K-1 at room temperature after quantum correction, with possible ~60% elastic and ~40% inelastic contributions. The result revises previous NEMD predictions (400 - 2000 MW m-2 K-1) using empirical interatomic potentials. When a second GaN/AlN interface is brought close to the original interface, the TBC of the original interface can increase to 1000 MW m-2 K-1, and this value gradually decreases with increasing distance between the two interfaces. When multiple interfaces are introduced in proximity, the first interface's TBC can be further enhanced to above 1150 MW m-2 K-1. After comparing double interfaces, superlattices, and random multilayers, it is concluded that such enhancement of TBC is not caused by the emergence of coherent superlattice modes but rather by the ballistic transport of the original bulk phonon modes of each material. Additionally, a "critical separation distance" (l_cs) is defined as the threshold beyond which the two interfaces no longer influence each other and behave independently. l_cs is determined by the mean free path of the phonons filtered by the interfaces that transport in the middle layer between two interfaces. Our findings provide insights into the thermal transport mechanisms that may aid the design of future electronic devices.

cond-mat.mes-hall↗

arXiv2025

Empirical Performance Evaluation of Lane Keeping Assist on Modern Production Vehicles

Leveraging a newly released open dataset of Lane Keeping Assist (LKA) systems from production vehicles, this paper presents the first comprehensive empirical analysis of real-world LKA performance. Our study yields three key findings: (i) LKA failures can be systematically categorized into perception, planning, and control errors. We present representative examples of each failure mode through in-depth analysis of LKA-related CAN signals, enabling both justification of the failure mechanisms and diagnosis of when and where each module begins to degrade; (ii) LKA systems tend to follow a fixed lane-centering strategy, often resulting in outward drift that increases linearly with road curvature, whereas human drivers proactively steer slightly inward on similar curved segments; (iii) We provide the first statistical summary and distribution analysis of environmental and road conditions under LKA failures, identifying with statistical significance that faded lane markings, low pavement laneline contrast, and sharp curvature are the most dominant individual factors, along with critical combinations that substantially increase failure likelihood. Building on these insights, we propose a theoretical model that integrates road geometry, speed limits, and LKA steering capability to inform infrastructure design. Additionally, we develop a machine learning-based model to assess roadway readiness for LKA deployment, offering practical tools for safer infrastructure planning, especially in rural areas. This work highlights key limitations of current LKA systems and supports the advancement of safer and more reliable autonomous driving technologies.

cs.RO↗

arXiv2025

Bridging Human Oversight and Black-box Driver Assistance: Vision-Language Models for Predictive Alerting in Lane Keeping Assist Systems

Lane Keeping Assist systems, while increasingly prevalent, often suffer from unpredictable real-world failures, largely due to their opaque, black-box nature, which limits driver anticipation and trust. To bridge the gap between automated assistance and effective human oversight, we present LKAlert, a novel supervisory alert system that leverages VLM to forecast potential LKA risk 1-3 seconds in advance. LKAlert processes dash-cam video and CAN data, integrating surrogate lane segmentation features from a parallel interpretable model as automated guiding attention. Unlike traditional binary classifiers, LKAlert issues both predictive alert and concise natural language explanation, enhancing driver situational awareness and trust. To support the development and evaluation of such systems, we introduce OpenLKA-Alert, the first benchmark dataset designed for predictive and explainable LKA failure warnings. It contains synchronized multimodal inputs and human-authored justifications across annotated temporal windows. We further contribute a generalizable methodological framework for VLM-based black-box behavior prediction, combining surrogate feature guidance with LoRA. This framework enables VLM to reason over structured visual context without altering its vision backbone, making it broadly applicable to other complex, opaque systems requiring interpretable oversight. Empirical results correctly predicts upcoming LKA failures with 69.8% accuracy and a 58.6\% F1-score. The system also generates high-quality textual explanations for drivers (71.7 ROUGE-L) and operates efficiently at approximately 2 Hz, confirming its suitability for real-time, in-vehicle use. Our findings establish LKAlert as a practical solution for enhancing the safety and usability of current ADAS and offer a scalable paradigm for applying VLMs to human-centered supervision of black-box automation.

cs.RO↗

arXiv2025

MobA: Multifaceted Memory-Enhanced Adaptive Planning for Efficient Mobile Task Automation

Existing Multimodal Large Language Model (MLLM)-based agents face significant challenges in handling complex GUI (Graphical User Interface) interactions on devices. These challenges arise from the dynamic and structured nature of GUI environments, which integrate text, images, and spatial relationships, as well as the variability in action spaces across different pages and tasks. To address these limitations, we propose MobA, a novel MLLM-based mobile assistant system. MobA introduces an adaptive planning module that incorporates a reflection mechanism for error recovery and dynamically adjusts plans to align with the real environment contexts and action module's execution capacity. Additionally, a multifaceted memory module provides comprehensive memory support to enhance adaptability and efficiency. We also present MobBench, a dataset designed for complex mobile interactions. Experimental results on MobBench and AndroidArena demonstrate MobA's ability to handle dynamic GUI environments and perform complex mobile tasks.

cs.MA↗

arXiv2025

AI-CDA4All: Democratizing Cooperative Autonomous Driving for All Drivers via Affordable Dash-cam Hardware and Open-source AI Software

As transportation technology advances, the demand for connected vehicle infrastructure has greatly increased to improve their efficiency and safety. One area of advancement, Cooperative Driving Automation (CDA) still relies on expensive autonomy sensors or connectivity units and are not interoperable across existing market car makes/models, limiting its scalability on public roads. To fill these gaps, this paper presents a novel approach to democratizing CDA technology, it leverages low-cost, commercially available edge devices such as vehicle dash-cams and open-source software to make the technology accessible and scalable to be used in transportation infrastructure and broader public domains. This study also investigates the feasibility of utilizing cost-effective communication protocols based on LTE and WiFi. These technologies enable lightweight Vehicle-to-Everything (V2X) communications, facilitating real-time data exchange between vehicles and infrastructure. Our research and development efforts are aligned with industrial standards to ensure compatibility and future integration into existing transportation ecosystems. By prioritizing infrastructure-oriented applications, such as improved traffic flow management, this approach seeks to deliver tangible societal benefits without directly competing with vehicle OEMs. As recent advancement of Generative AI (GenAI), there is no standardized integration of GenAI technologies into open-source CDAs, as the current trends of muiltimodal large language models gain popularity, we demonstrated a feasible locally deployed edge LLM models can enhance driving experience while preserving privacy and security compared to cloud-connected solutions. The proposed system underscores the potential of low-cost, scalable solutions in advancing CDA functionality, paving the way for smarter, safer, and more inclusive transportation networks.

eess.SY↗

arXiv2025

Gemini: A Family of Highly Capable Multimodal Models

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

cs.CL↗

arXiv2025

HRScene: How Far Are VLMs from Effective High-Resolution Image Understanding?

High-resolution image (HRI) understanding aims to process images with a large number of pixels, such as pathological images and agricultural aerial images, both of which can exceed 1 million pixels. Vision Large Language Models (VLMs) can allegedly handle HRIs, however, there is a lack of a comprehensive benchmark for VLMs to evaluate HRI understanding. To address this gap, we introduce HRScene, a novel unified benchmark for HRI understanding with rich scenes. HRScene incorporates 25 real-world datasets and 2 synthetic diagnostic datasets with resolutions ranging from 1,024 $\times$ 1,024 to 35,503 $\times$ 26,627. HRScene is collected and re-annotated by 10 graduate-level annotators, covering 25 scenarios, ranging from microscopic to radiology images, street views, long-range pictures, and telescope images. It includes HRIs of real-world objects, scanned documents, and composite multi-image. The two diagnostic evaluation datasets are synthesized by combining the target image with the gold answer and distracting images in different orders, assessing how well models utilize regions in HRI. We conduct extensive experiments involving 28 VLMs, including Gemini 2.0 Flash and GPT-4o. Experiments on HRScene show that current VLMs achieve an average accuracy of around 50% on real-world tasks, revealing significant gaps in HRI understanding. Results on synthetic datasets reveal that VLMs struggle to effectively utilize HRI regions, showing significant Regional Divergence and lost-in-middle, shedding light on future research.

cs.CL↗

arXiv2025

Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use

Reinforcement learning has been shown to improve the performance of large language models. However, traditional approaches like RLHF or RLAIF treat the problem as single-step. As focus shifts toward more complex reasoning and agentic tasks, language models must take multiple steps of text generation, reasoning and environment interaction before generating a solution. We propose a synthetic data generation and RL methodology targeting multi-step optimization scenarios. This approach, called Step-Wise Reinforcement Learning (SWiRL), iteratively generates multi-step reasoning and tool use data, and then learns from that data. It employs a simple step-wise decomposition that breaks each multi-step trajectory into multiple sub-trajectories corresponding to each action by the original model. It then applies synthetic data filtering and RL optimization on these sub-trajectories. We evaluated SWiRL on a number of multi-step tool use, question answering, and mathematical reasoning tasks. Our experiments show that SWiRL outperforms baseline approaches by 21.5%, 12.3%, 14.8%, 11.1%, and 15.3% in relative accuracy on GSM8K, HotPotQA, CofCA, MuSiQue, and BeerQA, respectively. Excitingly, the approach exhibits generalization across tasks: for example, training only on HotPotQA (text question-answering) improves zero-shot performance on GSM8K (a math dataset) by a relative 16.9%.

cs.AI↗

arXiv2025

Intelligent Attacks and Defense Methods in Federated Learning-enabled Energy-Efficient Wireless Networks

Federated learning (FL) is a promising technique for learning-based functions in wireless networks, thanks to its distributed implementation capability. On the other hand, distributed learning may increase the risk of exposure to malicious attacks where attacks on a local model may spread to other models by parameter exchange. Meanwhile, such attacks can be hard to detect due to the dynamic wireless environment, especially considering local models can be heterogeneous with non-independent and identically distributed (non-IID) data. Therefore, it is critical to evaluate the effect of malicious attacks and develop advanced defense techniques for FL-enabled wireless networks. In this work, we introduce a federated deep reinforcement learning-based cell sleep control scenario that enhances the energy efficiency of the network. We propose multiple intelligent attacks targeting the learning-based approach and we propose defense methods to mitigate such attacks. In particular, we have designed two attack models, generative adversarial network (GAN)-enhanced model poisoning attack and regularization-based model poisoning attack. As a counteraction, we have proposed two defense schemes, autoencoder-based defense, and knowledge distillation (KD)-enabled defense. The autoencoder-based defense method leverages an autoencoder to identify the malicious participants and only aggregate the parameters of benign local models during the global aggregation, while KD-based defense protects the model from attacks by controlling the knowledge transferred between the global model and local models.

cs.LG↗

arXiv2025

Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training

Large language models (LLMs) exhibit remarkable multilingual capabilities despite the extreme language imbalance in the pre-training data. In this paper, we closely examine the reasons behind this phenomenon, focusing on the pre-training corpus. We find that the existence of code-switching, alternating between different languages within a context, is key to multilingual capabilities. We conduct an analysis to investigate code-switching in the pre-training corpus, examining its presence and categorizing it into four types within two quadrants. We then assess its impact on multilingual performance. These types of code-switching data are unbalanced in proportions and demonstrate different effects on facilitating language transfer. To better explore the power of code-switching for language alignment during pre-training, we investigate the strategy of synthetic code-switching. We continuously scale up the synthetic code-switching data and observe remarkable improvements in both benchmarks and representation space. Extensive experiments indicate that incorporating synthetic code-switching data enables better language alignment and generalizes well to high, medium, and low-resource languages with pre-training corpora of varying qualities.

cs.CL↗

arXiv2025

Numerical analysis of American option pricing in a two-asset jump-diffusion model

This paper addresses an important gap in rigorous numerical treatments for pricing American options under correlated two-asset jump-diffusion models using the viscosity solution framework, with a particular focus on the Merton model. The pricing of these options is governed by complex two-dimensional (2-D) variational inequalities that incorporate cross-derivative terms and nonlocal integro-differential terms due to the presence of jumps. Existing numerical methods, primarily based on finite differences, often struggle with preserving monotonicity in the approximation of cross-derivatives, a key requirement for ensuring convergence to the viscosity solution. In addition, these methods face challenges in accurately discretizing 2-D jump integrals. We introduce a novel approach to effectively tackle the aforementioned variational inequalities while seamlessly handling cross-derivative terms and nonlocal integro-differential terms through an efficient and straightforward-to-implement monotone integration scheme. Within each timestep, our approach explicitly enforces the inequality constraint, resulting in a 2-D Partial Integro-Differential Equation (PIDE) to solve. Its solution is expressed as a 2-D convolution integral involving the Green's function of the PIDE. We derive an infinite series representation of this Green's function, where each term is non-negative and computable. This facilitates the numerical approximation of the PIDE solution through a monotone integration method. To enhance efficiency, we develop an implementation of this monotone scheme via FFTs, exploiting the Toeplitz matrix structure. The proposed method is proved to be both $\ell_{\infty} $-stable and consistent in the viscosity sense, ensuring its convergence to the viscosity solution of the variational inequality. Extensive numerical results validate the effectiveness and robustness of our approach.

q-fin.CP↗

arXiv2025

TKwinFormer: Top k Window Attention in Vision Transformers for Feature Matching

Local feature matching remains a challenging task, primarily due to difficulties in matching sparse keypoints and low-texture regions. The key to solving this problem lies in effectively and accurately integrating global and local information. To achieve this goal, we introduce an innovative local feature matching method called TKwinFormer. Our approach employs a multi-stage matching strategy to optimize the efficiency of information interaction. Furthermore, we propose a novel attention mechanism called Top K Window Attention, which facilitates global information interaction through window tokens prior to patch-level matching, resulting in improved matching accuracy. Additionally, we design an attention block to enhance attention between channels. Experimental results demonstrate that TKwinFormer outperforms state-of-the-art methods on various benchmarks. Code is available at: https://github.com/LiaoYun0x0/TKwinFormer.

eess.IV↗

arXiv2025

UFM: Unified Feature Matching Pre-training with Multi-Modal Image Assistants

Image feature matching, a foundational task in computer vision, remains challenging for multimodal image applications, often necessitating intricate training on specific datasets. In this paper, we introduce a Unified Feature Matching pre-trained model (UFM) designed to address feature matching challenges across a wide spectrum of modal images. We present Multimodal Image Assistant (MIA) transformers, finely tunable structures adept at handling diverse feature matching problems. UFM exhibits versatility in addressing both feature matching tasks within the same modal and those across different modals. Additionally, we propose a data augmentation algorithm and a staged pre-training strategy to effectively tackle challenges arising from sparse data in specific modals and imbalanced modal datasets. Experimental results demonstrate that UFM excels in generalization and performance across various feature matching tasks. The code will be released at:https://github.com/LiaoYun0x0/UFM.

cs.CV↗

arXiv2025

Change3D: Revisiting Change Detection and Captioning from A Video Modeling Perspective

In this paper, we present Change3D, a framework that reconceptualizes the change detection and captioning tasks through video modeling. Recent methods have achieved remarkable success by regarding each pair of bi-temporal images as separate frames. They employ a shared-weight image encoder to extract spatial features and then use a change extractor to capture differences between the two images. However, image feature encoding, being a task-agnostic process, cannot attend to changed regions effectively. Furthermore, different change extractors designed for various change detection and captioning tasks make it difficult to have a unified framework. To tackle these challenges, Change3D regards the bi-temporal images as comprising two frames akin to a tiny video. By integrating learnable perception frames between the bi-temporal images, a video encoder enables the perception frames to interact with the images directly and perceive their differences. Therefore, we can get rid of the intricate change extractors, providing a unified framework for different change detection and captioning tasks. We verify Change3D on multiple tasks, encompassing change detection (including binary change detection, semantic change detection, and building damage assessment) and change captioning, across eight standard benchmarks. Without bells and whistles, this simple yet effective framework can achieve superior performance with an ultra-light video model comprising only ~6%-13% of the parameters and ~8%-34% of the FLOPs compared to state-of-the-art methods. We hope that Change3D could be an alternative to 2D-based models and facilitate future research.

cs.CV↗

arXiv2025

Solar Radio Burst Detection Based on Deformable DETR

Solar radio bursts (SRBs) detection is crucial for solar physics research and space weather forecasting. The main challenges faced are noise interference in the spectrum and the diversity of SRBs. However, most research focuses on classifying whether SRBs exist or detecting a single type of SRBs. Existing detection models exhibit deficiencies in the accuracy of SRBs detection. Moreover, existing detection models cannot effectively handle background noise interference in solar radio spectrograms and the significant scale variations among different burst types. This paper proposes a high-performance detection model for solar radio bursts (SRBs) based on Deformable DETR (DEtection TRansformers) called DETR4SBRs. Firstly, this study designed a scale sensitive attention (SSA) module better to address the scale variations of SRBs. Subsequently, this study introduced collaborative hybrid auxiliary training to mitigate the positive-negative sample imbalance issue in Deformable DETR. The experimental results demonstrate that the proposed model achieves a mAP@50 of 83.5% and a recall rate of 99.4% on the SRBs dataset. Additionally, the model exhibits excellent noise-robust performance and can efficiently detect and locate Type II, III, IV, and V SRBs. The model proposed in this study provides robust support for preliminary solar radio burst data processing and has significant implications for space weather forecasting. The source code and data are available on the https://github.com/onewangqianqian/SSA-Co-Deformable-DETR.git and archived on Zenodo.

astro-ph.IM↗

arXiv2025

Computationally and Sample Efficient Safe Reinforcement Learning Using Adaptive Conformal Prediction

Safety is a critical concern in learning-enabled autonomous systems especially when deploying these systems in real-world scenarios. An important challenge is accurately quantifying the uncertainty of unknown models to generate provably safe control policies that facilitate the gathering of informative data, thereby achieving both safe and optimal policies. Additionally, the selection of the data-driven model can significantly impact both the real-time implementation and the uncertainty quantification process. In this paper, we propose a provably sample efficient episodic safe learning framework that remains robust across various model choices with quantified uncertainty for online control tasks. Specifically, we first employ Quadrature Fourier Features (QFF) for kernel function approximation of Gaussian Processes (GPs) to enable efficient approximation of unknown dynamics. Then the Adaptive Conformal Prediction (ACP) is used to quantify the uncertainty from online observations and combined with the Control Barrier Functions (CBF) to characterize the uncertainty-aware safe control constraints under learned dynamics. Finally, an optimism-based exploration strategy is integrated with ACP-based CBFs for safe exploration and near-optimal safe nonlinear control. Theoretical proofs and simulation results are provided to demonstrate the effectiveness and efficiency of the proposed framework.

cs.RO↗