Searcharxiv⌕ Search

arXiv subjects

Qi Zhang

Publications and source records attributed to Qi Zhang.

At least 361 records · Page 20Linked to original sources
arXiv2025

A Dual-Modulation Framework for RGB-T Crowd Counting via Spatially Modulated Attention and Adaptive Fusion

Accurate RGB-Thermal (RGB-T) crowd counting is crucial for public safety in challenging conditions. While recent Transformer-based methods excel at capturing global context, their inherent lack of spatial inductive bias causes attention to spread to irrelevant background regions, compromising crowd localization precision. Furthermore, effectively bridging the gap between these distinct modalities remains a major hurdle. To tackle this, we propose the Dual Modulation Framework, comprising two modules: Spatially Modulated Attention (SMA), which improves crowd localization by using a learnable Spatial Decay Mask to penalize attention between distant tokens and prevent focus from spreading to the background; and Adaptive Fusion Modulation (AFM), which implements a dynamic gating mechanism to prioritize the most reliable modality for adaptive cross-modal fusion. Extensive experiments on RGB-T crowd counting datasets demonstrate the superior performance of our method compared to previous works. Code available at https://github.com/Cht2924/RGBT-Crowd-Counting.

cs.CV↗
arXiv2025

Certifiably Optimal Doppler Positioning using Opportunistic LEO Satellites

To provide backup and augmentation to global navigation satellite system (GNSS), Doppler shift from Low Earth Orbit (LEO) satellites can be employed as signals of opportunity (SOP) for position, navigation and timing (PNT). Since the Doppler positioning problem is non-convex, local searching methods may produce two types of estimates: a global optimum without notice or a local optimum given an inexact initial estimate. As exact initialization is unavailable in some unknown environments, a guaranteed global optimization method in no need of initialization becomes necessary. To achieve this goal, we propose a certifiably optimal LEO Doppler positioning method by utilizing convex optimization. In this paper, the certifiable positioning method is implemented through a graduated weight approximation (GWA) algorithm and semidefinite programming (SDP) relaxation. To guarantee the optimality, we derive the necessary conditions for optimality in ideal noiseless cases and sufficient noise bounds conditions in noisy cases. Simulation and real tests are conducted to evaluate the effectiveness and robustness of the proposed method. Specially, the real test using Iridium-NEXT satellites shows that the proposed method estimates an certifiably optimal solution with an 3D positioning error of 140 m without initial estimates while Gauss-Newton and Dog-Leg are trapped in local optima when the initial point is equal or larger than 1000 km away from the ground truth. Moreover, the certifiable estimation can also be used as initialization in local searching methods to lower down the 3D positioning error to 130 m.

cs.RO↗
arXiv2025

Highly Efficient Direct Analytics on Semantic-aware Time Series Data Compression

Semantic communication has emerged as a promising paradigm to tackle the challenges of massive growing data traffic and sustainable data communication. It shifts the focus from data fidelity to goal-oriented or task-oriented semantic transmission. While deep learning-based methods are commonly used for semantic encoding and decoding, they struggle with the sequential nature of time series data and high computation cost, particularly in resource-constrained IoT environments. Data compression plays a crucial role in reducing transmission and storage costs, yet traditional data compression methods fall short of the demands of goal-oriented communication systems. In this paper, we propose a novel method for direct analytics on time series data compressed by the SHRINK compression algorithm. Through experimentation using outlier detection as a case study, we show that our method outperforms baselines running on uncompressed data in multiple cases, with merely 1% difference in the worst case. Additionally, it achieves four times lower runtime on average and accesses approximately 10% of the data volume, which enables edge analytics with limited storage and computation power. These results demonstrate that our approach offers reliable, high-speed outlier detection analytics for diverse IoT applications while extracting semantics from time-series data, achieving high compression, and reducing data transmission.

cs.LG↗
arXiv2025

TASAM: Terrain-and-Aware Segment Anything Model for Temporal-Scale Remote Sensing Segmentation

Segment Anything Model (SAM) has demonstrated impressive zero-shot segmentation capabilities across natural image domains, but it struggles to generalize to the unique challenges of remote sensing data, such as complex terrain, multi-scale objects, and temporal dynamics. In this paper, we introduce TASAM, a terrain and temporally-aware extension of SAM designed specifically for high-resolution remote sensing image segmentation. TASAM integrates three lightweight yet effective modules: a terrain-aware adapter that injects elevation priors, a temporal prompt generator that captures land-cover changes over time, and a multi-scale fusion strategy that enhances fine-grained object delineation. Without retraining the SAM backbone, our approach achieves substantial performance gains across three remote sensing benchmarks-LoveDA, iSAID, and WHU-CD-outperforming both zero-shot SAM and task-specific models with minimal computational overhead. Our results highlight the value of domain-adaptive augmentation for foundation models and offer a scalable path toward more robust geospatial segmentation.

cs.CV↗
arXiv2025

A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learning

Robotic real-world reinforcement learning (RL) with vision-language-action (VLA) models is bottlenecked by sparse, handcrafted rewards and inefficient exploration. We introduce VLAC, a general process reward model built upon InternVL and trained on large scale heterogeneous datasets. Given pairwise observations and a language goal, it outputs dense progress delta and done signal, eliminating task-specific reward engineering, and supports one-shot in-context transfer to unseen tasks and environments. VLAC is trained on vision-language datasets to strengthen perception, dialogic and reasoning capabilities, together with robot and human trajectories data that ground action generation and progress estimation, and additionally strengthened to reject irrelevant prompts as well as detect regression or stagnation by constructing large numbers of negative and semantically mismatched samples. With prompt control, a single VLAC model alternately generating reward and action tokens, unifying critic and policy. Deployed inside an asynchronous real-world RL loop, we layer a graded human-in-the-loop protocol (offline demonstration replay, return and explore, human guided explore) that accelerates exploration and stabilizes early learning. Across four distinct real-world manipulation tasks, VLAC lifts success rates from about 30\% to about 90\% within 200 real-world interaction episodes; incorporating human-in-the-loop interventions yields a further 50% improvement in sample efficiency and achieves up to 100% final success.

cs.RO↗
arXiv2025

CUFG: Curriculum Unlearning Guided by the Forgetting Gradient

As privacy and security take center stage in AI, machine unlearning, the ability to erase specific knowledge from models, has garnered increasing attention. However, existing methods overly prioritize efficiency and aggressive forgetting, which introduces notable limitations. In particular, radical interventions like gradient ascent, influence functions, and random label noise can destabilize model weights, leading to collapse and reduced reliability. To address this, we propose CUFG (Curriculum Unlearning via Forgetting Gradients), a novel framework that enhances the stability of approximate unlearning through innovations in both forgetting mechanisms and data scheduling strategies. Specifically, CUFG integrates a new gradient corrector guided by forgetting gradients for fine-tuning-based unlearning and a curriculum unlearning paradigm that progressively forgets from easy to hard. These innovations narrow the gap with the gold-standard Retrain method by enabling more stable and progressive unlearning, thereby improving both effectiveness and reliability. Furthermore, we believe that the concept of curriculum unlearning has substantial research potential and offers forward-looking insights for the development of the MU field. Extensive experiments across various forgetting scenarios validate the rationale and effectiveness of our approach and CUFG. Codes are available at https://anonymous.4open.science/r/CUFG-6375.

cs.LG↗
arXiv2025

Querying Climate Knowledge: Semantic Retrieval for Scientific Discovery

The growing complexity and volume of climate science literature make it increasingly difficult for researchers to find relevant information across models, datasets, regions, and variables. This paper introduces a domain-specific Knowledge Graph (KG) built from climate publications and broader scientific texts, aimed at improving how climate knowledge is accessed and used. Unlike keyword based search, our KG supports structured, semantic queries that help researchers discover precise connections such as which models have been validated in specific regions or which datasets are commonly used with certain teleconnection patterns. We demonstrate how the KG answers such questions using Cypher queries, and outline its integration with large language models in RAG systems to improve transparency and reliability in climate-related question answering. This work moves beyond KG construction to show its real world value for climate researchers, model developers, and others who rely on accurate, contextual scientific information.

cs.CL↗
arXiv2025

Enhancing understanding and clinical applications of cerebral autoregulation: A novel integrated numerical framework

Cerebral autoregulation (CA) is a fundamental mechanism that modulates cerebrovascular resistance, primarily by regulating the diameter of small cerebral vessels to maintain stable cerebral blood flow (CBF) in response to fluctuations in systemic arterial pressure. However, the clinical understanding of CA remains limited due to the intricate structure of the cerebral vasculature and the challenges in accurately quantifying the hemodynamic and physiological parameters that govern this autoregulatory process. Method: In this study, we introduced a novel numerical algorithm that employs three partial differential equations and one ordinary differential equation to capture both the spatial and temporal distributions of key CA-driving factors, including the arterial pressure (P) and the partial pressures of oxygen (PO_2) and carbon dioxide (PCO_2) within the cerebral vasculature, together with a Windkessel model in turn to regulate the CBF based on the calculated P, PO_2, and PCO_2. This algorithm was sequentially integrated with our previously developed personalized 0D-1D multi-dimensional model to account for the patient-specific effects. Results: The integrated framework was rigorously validated using two independent datasets, demonstrating its high reliability and accuracy in capturing the regulatory effects of CA on CBF across a range of physiological conditions. Conclusion: This work significantly advances our understanding of CA and provides a promising foundation for developing hemodynamic-based therapeutic strategies aimed at improving clinical outcomes in patients with cerebrovascular disorders.

physics.med-ph↗
arXiv2025

A Conflicts-free, Speed-lossless KAN-based Reinforcement Learning Decision System for Interactive Driving in Roundabouts

Safety and efficiency are crucial for autonomous driving in roundabouts, especially mixed traffic with both autonomous vehicles (AVs) and human-driven vehicles. This paper presents a learning-based algorithm that promotes safe and efficient driving across varying roundabout traffic conditions. A deep Q-learning network is used to learn optimal strategies in complex multi-vehicle roundabout scenarios, while a Kolmogorov-Arnold Network (KAN) improves the AVs' environmental understanding. To further enhance safety, an action inspector filters unsafe actions, and a route planner optimizes driving efficiency. Moreover, model predictive control ensures stability and precision in execution. Experimental results demonstrate that the proposed system consistently outperforms state-of-the-art methods, achieving fewer collisions, reduced travel time, and stable training with smooth reward convergence.

cs.RO↗
arXiv2025

MagicGUI: A Foundational Mobile GUI Agent with Scalable Data Pipeline and Reinforcement Fine-tuning

This paper presents MagicGUI, a foundational mobile GUI agent designed to address critical challenges in perception, grounding, and reasoning within real-world mobile GUI environments. The framework is underpinned by following six key components: (1) a comprehensive and accurate dataset, constructed via the scalable GUI Data Pipeline, which aggregates the largest and most diverse GUI-centric multimodal data to date from open-source repositories, automated crawling, and targeted manual annotation; (2) enhanced perception and grounding capabilities, facilitating fine-grained multimodal alignment for UI element referencing, grounding, and screen comprehension; (3) a comprehensive and unified action space, encompassing both fundamental UI operations and complex interactive intents to support human-agent interactions; (4) planning-oriented reasoning mechanisms that enable the model to decompose complex user instructions into sequential actions with explicit intermediate meta-paln reasoning; (5) an iterative two-stage training procedure, combining large-scale continue pre-training on 7.8M samples with reinforcement fine-tuning utilizing a spatially enhanced composite reward and dual filtering strategy; and (6) competitive performance on both the proprietary Magic-RICH benchmark and over a dozen public benchmarks, achieving superior performance across GUI perception and agent tasks, while demonstrating robust generalization and real-world deployment potential in practical mobile GUI scenarios, as detailed in Figure 1.

cs.HC↗
arXiv2025

Entropy production rate and time-reversibility for general jump diffusions on $\mathbb{R}^n$

This paper investigates the entropy production rate and time-reversibility for general jump diffusions (Lévy processes) on $\mathbb{R}^n$. We first formulate the entropy production rate and explore its associated thermodynamic relations for jump diffusions. Subsequently, we derive the entropy production rate using the relative entropy between the forward and time-reversed path measures for stationary jump diffusions via the Girsanov transform. Furthermore, we establish the equivalence among time-reversibility, zero entropy production rate, detailed balance condition, and the gradient structure for stationary jump diffusions.

math.PR↗
arXiv2025

AgentGym-RL: Training LLM Agents for Long-Horizon Decision Making through Multi-Turn Reinforcement Learning

Developing autonomous LLM agents capable of making a series of intelligent decisions to solve complex, real-world tasks is a fast-evolving frontier. Like human cognitive development, agents are expected to acquire knowledge and skills through exploration and interaction with the environment. Despite advances, the community still lacks a unified, interactive reinforcement learning (RL) framework that can effectively train such agents from scratch -- without relying on supervised fine-tuning (SFT) -- across diverse and realistic environments. To bridge this gap, we introduce AgentGym-RL, a new framework to train LLM agents for multi-turn interactive decision-making through RL. The framework features a modular and decoupled architecture, ensuring high flexibility and extensibility. It encompasses a wide variety of real-world scenarios, and supports mainstream RL algorithms. Furthermore, we propose ScalingInter-RL, a training approach designed for exploration-exploitation balance and stable RL optimization. In early stages, it emphasizes exploitation by restricting the number of interactions, and gradually shifts towards exploration with larger horizons to encourage diverse problem-solving strategies. In this way, the agent develops more diverse behaviors and is less prone to collapse under long horizons. We perform extensive experiments to validate the stability and effectiveness of both the AgentGym-RL framework and the ScalingInter-RL approach. Our agents match or surpass commercial models on 27 tasks across diverse environments. We offer key insights and will open-source the complete AgentGym-RL framework -- including code and datasets -- to empower the research community in developing the next generation of intelligent agents.

cs.LG↗
arXiv2025

SolarSeer: Ultrafast and accurate 24-hour solar irradiance forecasts outperforming numerical weather prediction across the USA

Accurate 24-hour solar irradiance forecasting is essential for the safe and economic operation of solar photovoltaic systems. Traditional numerical weather prediction (NWP) models represent the state-of-the-art in forecasting performance but rely on computationally costly data assimilation and solving complicated partial differential equations (PDEs) that simulate atmospheric physics. Here, we introduce SolarSeer, an end-to-end large artificial intelligence (AI) model for solar irradiance forecasting across the Contiguous United States (CONUS). SolarSeer is designed to directly map the historical satellite observations to future forecasts, eliminating the computational overhead of data assimilation and PDEs solving. This efficiency allows SolarSeer to operate over 1,500 times faster than traditional NWP, generating 24-hour cloud cover and solar irradiance forecasts for the CONUS at 5-kilometer resolution in under 3 seconds. Compared with the state-of-the-art NWP in the CONUS, i.e., High-Resolution Rapid Refresh (HRRR), SolarSeer significantly reduces the root mean squared error of solar irradiance forecasting by 27.28% in reanalysis data and 15.35% across 1,800 stations. SolarSeer also effectively captures solar irradiance fluctuations and significantly enhances the first-order irradiance difference forecasting accuracy. SolarSeer's ultrafast, accurate 24-hour solar irradiance forecasts provide strong support for the transition to sustainable, net-zero energy systems.

cs.LG↗
arXiv2025

Distill Visual Chart Reasoning Ability from LLMs to MLLMs

Solving complex chart Q&A tasks requires advanced visual reasoning abilities in multimodal large language models (MLLMs), including recognizing key information from visual inputs and conducting reasoning over it. While fine-tuning MLLMs for reasoning is critical, collecting and annotating charts and questions is expensive, hard to scale, and often results in low-quality annotations. To address this, we propose Code-as-Intermediary Translation (CIT), a cost-effective, efficient and scalable data synthesis method for distilling visual reasoning abilities from LLMs to MLLMs. The code serves as an intermediary that translates visual chart representations into textual representations, enabling language models to understand cross-modal information and generate reasoning chains accordingly. In this way, we can employ text-based synthesizing techniques to expand chart-plotting code and generate high-quality Q&A pairs for training models. This produces ReachQA, a dataset containing 3k reasoning-intensive charts and 20k Q&A pairs to enhance both recognition and reasoning abilities of MLLMs. Experiments show that models fine-tuned with ReachQA not only perform well on chart-related tasks but also show performance gains on general reasoning benchmarks. The code and dataset are publicly available at https://github.com/hewei2001/ReachQA.

cs.CL↗
arXiv2025

LLMEval-Med: A Real-world Clinical Benchmark for Medical LLMs with Physician Validation

Evaluating large language models (LLMs) in medicine is crucial because medical applications require high accuracy with little room for error. Current medical benchmarks have three main types: medical exam-based, comprehensive medical, and specialized assessments. However, these benchmarks have limitations in question design (mostly multiple-choice), data sources (often not derived from real clinical scenarios), and evaluation methods (poor assessment of complex reasoning). To address these issues, we present LLMEval-Med, a new benchmark covering five core medical areas, including 2,996 questions created from real-world electronic health records and expert-designed clinical scenarios. We also design an automated evaluation pipeline, incorporating expert-developed checklists into our LLM-as-Judge framework. Furthermore, our methodology validates machine scoring through human-machine agreement analysis, dynamically refining checklists and prompts based on expert feedback to ensure reliability. We evaluate 13 LLMs across three categories (specialized medical models, open-source models, and closed-source models) on LLMEval-Med, providing valuable insights for the safe and effective deployment of LLMs in medical domains. The dataset is released in https://github.com/llmeval/LLMEval-Med.

cs.CL↗
arXiv2025

Quenched Quantum Feature Maps

We propose a quantum feature mapping technique that leverages the quench dynamics of a quantum spin glass to extract complex data patterns at the quantum-advantage level for academic and industrial applications. We demonstrate that encoding a dataset information into disordered quantum many-body spin-glass problems, followed by a nonadiabatic evolution and feature extraction via measurements of expectation values, significantly enhances machine learning (ML) models. By analyzing the performance of our protocol over a range of evolution times, we empirically show that ML models benefit most from feature representations obtained in the fast coherent regime of a quantum annealer, particularly near the critical point of the quantum dynamics. We demonstrate the generalization of our technique by benchmarking on multiple high-dimensional datasets, involving over a hundred features, in applications including drug discovery and medical diagnostics. Moreover, we compare against a comprehensive suite of state-of-the-art classical ML models and show that our quantum feature maps can enhance the performance metrics of the baseline classical models up to 210%. Our work presents the first quantum ML demonstrations at the quantum-advantage level, bridging the gap between quantum supremacy and useful real-world academic and industrial applications.

quant-ph↗
arXiv2025

TL-Training: A Task-Feature-Based Framework for Training Large Language Models in Tool Use

Large language models (LLMs) achieve remarkable advancements by leveraging tools to interact with environments, a critical step toward generalized AI. However, the standard supervised fine-tuning (SFT) approach, which relies on large-scale datasets, often overlooks task-specific characteristics in tool use, leading to performance bottlenecks. To address this issue, we analyze three existing LLMs and uncover key insights: training data can inadvertently impede tool-use behavior, token importance is distributed unevenly, and errors in tool calls fall into a small set of categories. Building on these findings, we propose~\emph{TL-Training}, a task-feature-based framework that mitigates the effects of suboptimal training data, dynamically adjusts token weights to prioritize key tokens during SFT, and incorporates a robust reward mechanism tailored to error categories, optimized through proximal policy optimization. We validate TL-Training by training CodeLLaMA-2-7B and evaluating it on four open-source test sets. Our results demonstrate that the LLM trained by our method matches or surpasses both open- and closed-source LLMs in tool-use performance using only 1,217 training data points. Additionally, our method enhances robustness in noisy environments and improves general task performance, offering a scalable and efficient paradigm for tool-use training in LLMs. Code and data are available at https://github.com/Junjie-Ye/TL-Training.

cs.CL↗