arXiv subjects

Hao Zhou

Publications and source records attributed to Hao Zhou.

At least 127 records · Page 7Linked to original sources

arXiv2026

ADAS-TO: A Large-Scale Multimodal Naturalistic Dataset and Empirical Characterization of Human Takeovers during ADAS Engagement

Takeovers remain a key safety vulnerability in production ADAS, yet existing public resources rarely provide takeover-centered, real-world data. We present ADAS-TO, the first large-scale naturalistic dataset dedicated to ADAS-to-manual transitions, containing 15,659 takeover-centered 20s clips from 327 drivers across 22 vehicle brands. Each clip synchronizes front-view video with CAN logs. Takeovers are defined as ADAS ON $\rightarrow$ OFF transitions, with the primary trigger labeled as brake, steer, gas, mixed, or system disengagement. We further separate planned driver-initiated terminations (Ego) from forced takeovers (Non-ego) using a rule-based partition. While most events occur within conservative kinematic margins, we identify a long tail of 285 safety-critical cases. For these events, we combine kinematic screening with vision--language (VLM) annotation to attribute hazards and relate them to intervention dynamics. The resulting cross-modal analysis shows distinct kinematic signatures across traffic dynamics, infrastructure degradation, and adverse environments, and finds that in 59.3% of critical cases, actionable visual cues emerge at least 3s before takeover, supporting the potential for semantics-aware early warning beyond late-stage kinematic triggers. The dataset is publicly released at huggingface.co/datasets/HenryYHW/ADAS-TO-Sample.

cs.HC↗

arXiv2026

TaPD: Temporal-adaptive Progressive Distillation for Observation-Adaptive Trajectory Forecasting in Autonomous Driving

Trajectory prediction is essential for autonomous driving, enabling vehicles to anticipate the motion of surrounding agents to support safe planning. However, most existing predictors assume fixed-length histories and suffer substantial performance degradation when observations are variable or extremely short in real-world settings (e.g., due to occlusion or a limited sensing range). We propose TaPD (Temporal-adaptive Progressive Distillation), a unified plug-and-play framework for observation-adaptive trajectory forecasting under variable history lengths. TaPD comprises two cooperative modules: an Observation-Adaptive Forecaster (OAF) for future prediction and a Temporal Backfilling Module (TBM) for explicit reconstruction of the past. OAF is built on progressive knowledge distillation (PKD), which transfers motion pattern knowledge from long-horizon "teachers" to short-horizon "students" via hierarchical feature regression, enabling short observations to recover richer motion context. We further introduce a cosine-annealed distillation weighting scheme to balance forecasting supervision and feature alignment, improving optimization stability and cross-length consistency. For extremely short histories where implicit alignment is insufficient, TBM backfills missing historical segments conditioned on scene evolution, producing context-rich trajectories that strengthen PKD and thereby improve OAF. We employ a decoupled pretrain-reconstruct-finetune protocol to preserve real-motion priors while adapting to backfilled inputs. Extensive experiments on Argoverse 1 and Argoverse 2 show that TaPD consistently outperforms strong baselines across all observation lengths, delivers especially large gains under very short inputs, and improves other predictors (e.g., HiVT) in a plug-and-play manner. Code will be available at https://github.com/zhouhao94/TaPD.

cs.CV↗

arXiv2026

MoMa: A Modular Deep Learning Framework for Material Property Prediction

Deep learning methods for material property prediction have been widely explored to advance materials discovery. However, the prevailing pre-train then fine-tune paradigm often fails to address the inherent diversity and disparity of material tasks. To overcome these challenges, we introduce MoMa, a Modular framework for Materials that first trains specialized modules across a wide range of tasks and then adaptively composes synergistic modules tailored to each downstream scenario. Evaluation across 17 datasets demonstrates the superiority of MoMa, with a substantial 14% average improvement over the strongest baseline. Few-shot and continual learning experiments further highlight MoMa's potential for real-world applications. Pioneering a new paradigm of modular material learning, MoMa will be open-sourced to foster broader community collaboration.

cs.LG↗

arXiv2026

Quark Medical Alignment: A Holistic Multi-Dimensional Alignment and Collaborative Optimization Paradigm

While reinforcement learning for large language model alignment has progressed rapidly in recent years, transferring these paradigms to high-stakes medical question answering reveals a fundamental paradigm mismatch. Reinforcement Learning from Human Feedback relies on preference annotations that are prohibitively expensive and often fail to reflect the absolute correctness of medical facts. Reinforcement Learning from Verifiable Rewards lacks effective automatic verifiers and struggles to handle complex clinical contexts. Meanwhile, medical alignment requires the simultaneous optimization of correctness, safety, and compliance, yet multi-objective heterogeneous reward signals are prone to scale mismatch and optimization conflicts. To address these challenges, we propose a robust medical alignment paradigm. We first construct a holistic multi-dimensional medical alignment matrix that decomposes alignment objectives into four categories: fundamental capabilities, expert knowledge, online feedback, and format specifications. Within each category, we establish a closed loop of where observable metrics inform attributable diagnosis, which in turn drives optimizable rewards, thereby providing fine-grained, high-resolution supervision signals for subsequent iterative optimization. To resolve gradient domination and optimization instability problem caused by heterogeneous signals, we further propose a unified optimization mechanism. This mechanism employs Reference-Frozen Normalization to align reward scales and implements a Tri-Factor Adaptive Dynamic Weighting strategy to achieve collaborative optimization that is weakness-oriented, risk-prioritized, and redundancy-reducing. Experimental results demonstrate the effectiveness of our proposed paradigm in real-world medical scenario evaluations, establishing a new paradigm for complex alignment in vertical domains.

cs.AI↗

arXiv2026

Large Language Model-Assisted UAV Operations and Communications: A Multifaceted Survey and Tutorial

Uncrewed Aerial Vehicles (UAVs) are widely deployed across diverse applications due to their mobility and agility. Recent advances in Large Language Models (LLMs) offer a transformative opportunity to enhance UAV intelligence beyond conventional optimization-based and learning-based approaches. By integrating LLMs into UAV systems, advanced environmental understanding, swarm coordination, mobility optimization, and high-level task reasoning can be achieved, thereby allowing more adaptive and context-aware aerial operations. This survey systematically explores the intersection of LLMs and UAV technologies and proposes a unified framework that consolidates existing architectures, methodologies, and applications for UAVs. We first present a structured taxonomy of LLM adaptation techniques for UAVs, including pretraining, fine-tuning, Retrieval-Augmented Generation (RAG), and prompt engineering, along with key reasoning capabilities such as Chain-of-Thought (CoT) and In-Context Learning (ICL). We then examine LLM-assisted UAV communications and operations, covering navigation, mission planning, swarm control, safety, autonomy, and network management. After that, the survey further discusses Multimodal LLMs (MLLMs) for human-swarm interaction, perception-driven navigation, and collaborative control. Finally, we address ethical considerations, including bias, transparency, accountability, and Human-in-the-Loop (HITL) strategies, and outline future research directions. Overall, this work positions LLM-assisted UAVs as a foundation for intelligent and adaptive aerial systems.

cs.RO↗

arXiv2026

CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation

GPU kernel optimization is fundamental to modern deep learning but remains a highly specialized task requiring deep hardware expertise. Despite strong performance in general programming, large language models (LLMs) remain uncompetitive with compiler-based systems such as torch.compile for CUDA kernel generation. Existing CUDA code generation approaches either rely on training-free refinement or fine-tune models within fixed multi-turn execution-feedback loops, but both paradigms fail to fundamentally improve the model's intrinsic CUDA optimization ability, resulting in limited performance gains. We present CUDA Agent, a large-scale agentic reinforcement learning system that develops CUDA kernel expertise through three components: a scalable data synthesis pipeline, a skill-augmented CUDA development environment with automated verification and profiling to provide reliable reward signals, and reinforcement learning algorithmic techniques enabling stable training. CUDA Agent achieves state-of-the-art results on KernelBench, delivering 100\%, 100\%, and 92\% faster rate over torch.compile on KernelBench Level-1, Level-2, and Level-3 splits, outperforming the strongest proprietary models such as Claude Opus 4.5 and Gemini 3 Pro by about 40\% on the hardest Level-3 setting.

cs.LG↗

arXiv2026

Enhancing Large Language Models (LLMs) for Telecom using Dynamic Knowledge Graphs and Explainable Retrieval-Augmented Generation

Large language models (LLMs) have shown strong potential across a variety of tasks, but their application in the telecom field remains challenging due to domain complexity, evolving standards, and specialized terminology. Therefore, general-domain LLMs may struggle to provide accurate and reliable outputs in this context, leading to increased hallucinations and reduced utility in telecom operations.To address these limitations, this work introduces KG-RAG-a novel framework that integrates knowledge graphs (KGs) with retrieval-augmented generation (RAG) to enhance LLMs for telecom-specific tasks. In particular, the KG provides a structured representation of domain knowledge derived from telecom standards and technical documents, while RAG enables dynamic retrieval of relevant facts to ground the model's outputs. Such a combination improves factual accuracy, reduces hallucination, and ensures compliance with telecom specifications.Experimental results across benchmark datasets demonstrate that KG-RAG outperforms both LLM-only and standard RAG baselines, e.g., KG-RAG achieves an average accuracy improvement of 14.3% over RAG and 21.6% over LLM-only models. These results highlight KG-RAG's effectiveness in producing accurate, reliable, and explainable outputs in complex telecom scenarios.

cs.AI↗

arXiv2026

From Prompts to Protection: Large Language Model-Enabled In-Context Learning for Smart Public Safety UAV

A public safety Uncrewed Aerial Vehicle (UAV) enhances situational awareness during emergency response. Its agility, mobility optimization, and ability to establish Line-of-Sight (LoS) communication make it increasingly important for managing emergencies such as disaster response, search and rescue, and wildfire monitoring. Although Deep Reinforcement Learning (DRL) has been used to optimize UAV navigation and control, its high training complexity, low sample efficiency, and the simulation-to-reality gap limit its practicality in public safety applications. Recent advances in Large Language Models (LLMs) present a promising alternative. With strong reasoning and generalization abilities, LLMs can adapt to new tasks through In-Context Learning (ICL), enabling task adaptation via natural language prompts and example-based guidance without retraining. Deploying LLMs at the network edge, rather than in the cloud, further reduces latency and preserves data privacy, making them suitable for real-time, mission-critical public safety UAVs. This paper proposes integrating LLM-assisted ICL with public safety UAVs to address key functions such as path planning and velocity control in emergency response. We present a case study on data collection scheduling, demonstrating that the LLM-assisted ICL framework can significantly reduce packet loss compared to conventional approaches while also mitigating potential jailbreaking vulnerabilities. Finally, we discuss LLM optimizers and outline future research directions. The ICL framework enables adaptive, context-aware decision-making for public safety UAVs, offering a lightweight and efficient solution to enhance UAV autonomy and responsiveness in emergencies.

cs.AI↗

arXiv2026

FRSICL: LLM-Enabled In-Context Learning Flight Resource Allocation for Fresh Data Collection in UAV-Assisted Wildfire Monitoring

Uncrewed Aerial Vehicles (UAVs) play a vital role in public safety, especially in monitoring wildfires, where early detection reduces environmental impact. In UAV-Assisted Wildfire Monitoring (UAWM) systems, jointly optimizing the data collection schedule and UAV velocity is essential to minimize the average Age of Information (AoI) for sensory data. Deep Reinforcement Learning (DRL) has been used for this optimization, but its limitations-including low sampling efficiency, discrepancies between simulation and real-world conditions, and complex training make it unsuitable for time-critical applications such as wildfire monitoring. Recent advances in Large Language Models (LLMs) provide a promising alternative. With strong reasoning and generalization capabilities, LLMs can adapt to new tasks through In-Context Learning (ICL), which enables task adaptation using natural language prompts and example-based guidance without retraining. This paper proposes a novel online Flight Resource Allocation scheme based on LLM-Enabled In-Context Learning (FRSICL) to jointly optimize the data collection schedule and UAV velocity along the trajectory in real time, thereby asymptotically minimizing the average AoI across all ground sensors. Unlike DRL, FRSICL generates data collection schedules and velocities using natural language task descriptions and feedback from the environment, enabling dynamic decision-making without extensive retraining. Simulation results confirm the effectiveness of FRSICL compared to state-of-the-art baselines, namely Proximal Policy Optimization, Block Coordinate Descent, and Nearest Neighbor.

cs.AI↗

arXiv2026

ML-ECS: A Collaborative Multimodal Learning Framework for Edge-Cloud Synergies

Edge-cloud synergies provide a promising paradigm for privacy-preserving deployment of foundation models, where lightweight on-device models adapt to domain-specific data and cloud-hosted models coordinate knowledge sharing. However, in real-world edge environments, collaborative multimodal learning is challenged by modality heterogeneity (different modality combinations across domains) and model-structure heterogeneity (different modality-specific encoders/fusion modules. To address these issues, we propose ML-ECS, a collaborative multimodal learning framework that enables joint training between a server-based model and heterogeneous edge models. This framework consists of four components: (1) cross-modal contrastive learning (CCL) to align modality representations in a shared latent space, (2) adaptive multimodal tuning (AMT) to preserve domain-specific knowledge from local datasets, (3) modality-aware model aggregation (MMA) to robustly aggregate while mitigating noise caused by missing modalities, and (4) SLM-enhanced CCL (SE-CCL) to facilitate bidirectional knowledge transfer between cloud and edge. Experimental results on various multimodal tasks show that \pname consistently outperform state-of-the-art baselines under varying modality availability, achieving improvements of 5.44% to 12.08% in Rouge-LSum and improving both client- and server-side performance. In addition, by communicating only low-rank LoRA parameters and fused representations, ML-ECS achieves high communication efficiency, requiring only 0.65% of the total parameter volume.

cs.DC↗

arXiv2026

Exploring the Feasibility of Full-Body Muscle Activation Sensing with Insole Pressure Sensors

Muscle activation initiates contractions that drive human movement, and understanding it provides valuable insights for injury prevention and rehabilitation. Yet, sensing muscle activation is barely explored in the rapidly growing mobile health market. Traditional methods for muscle activation sensing rely on specialized electrodes, such as surface electromyography, making them impractical, especially for long-term usage. In this paper, we introduce Press2Muscle, the first system to unobtrusively infer muscle activation using insole pressure sensors. The key idea is to analyze foot pressure changes resulting from full-body muscle activation that drives movements. To handle variations in pressure signals due to differences in users' gait, weight, and movement styles, we propose a data-driven approach to dynamically adjust reliance on different foot regions and incorporate easily accessible biographical data to enhance Press2Muscle's generalization to unseen users. We conducted an extensive study with 30 users. Under a leave-one-user-out setting, Press2Muscle achieves a root mean square error of 0.025, marking a 19% improvement over a video-based counterpart. A robustness study validates Press2Muscle's ability to generalize across user demographics, footwear types, and walking surfaces. Additionally, we showcase muscle imbalance detection and muscle activation estimation under free-living settings with Press2Muscle, confirming the feasibility of muscle activation sensing using insole pressure sensors in real-world settings.

cs.HC↗

arXiv2026

The Missing Half: Unveiling Training-time Implicit Safety Risks Beyond Deployment

Safety risks of AI models have been widely studied at deployment time, such as jailbreak attacks that elicit harmful outputs. In contrast, safety risks emerging during training remain largely unexplored. Beyond explicit reward hacking that directly manipulates explicit reward functions in reinforcement learning, we study implicit training-time safety risks: harmful behaviors driven by a model's internal incentives and contextual background information. For example, during code-based reinforcement learning, a model may covertly manipulate logged accuracy for self-preservation. We present the first systematic study of this problem, introducing a taxonomy with five risk levels, ten fine-grained risk categories, and three incentive types. Extensive experiments reveal the prevalence and severity of these risks: notably, Llama-3.1-8B-Instruct exhibits risky behaviors in 74.4% of training runs when provided only with background information. We further analyze factors influencing these behaviors and demonstrate that implicit training-time risks also arise in multi-agent training settings. Our results identify an overlooked yet urgent safety challenge in training.

cs.CL↗

arXiv2026

ASTRA: Automated Synthesis of agentic Trajectories and Reinforcement Arenas

Large language models (LLMs) are increasingly used as tool-augmented agents for multi-step decision making, yet training robust tool-using agents remains challenging. Existing methods still require manual intervention, depend on non-verifiable simulated environments, rely exclusively on either supervised fine-tuning (SFT) or reinforcement learning (RL), and struggle with stable long-horizon, multi-turn learning. To address these challenges, we introduce ASTRA, a fully automated end-to-end framework for training tool-augmented language model agents via scalable data synthesis and verifiable reinforcement learning. ASTRA integrates two complementary components. First, a pipeline that leverages the static topology of tool-call graphs synthesizes diverse, structurally grounded trajectories, instilling broad and transferable tool-use competence. Second, an environment synthesis framework that captures the rich, compositional topology of human semantic reasoning converts decomposed question-answer traces into independent, code-executable, and rule-verifiable environments, enabling deterministic multi-turn RL. Based on this method, we develop a unified training methodology that integrates SFT with online RL using trajectory-level rewards to balance task completion and interaction efficiency. Experiments on multiple agentic tool-use benchmarks demonstrate that ASTRA-trained models achieve state-of-the-art performance at comparable scales, approaching closed-source systems while preserving core reasoning ability. We release the full pipelines, environments, and trained models at https://github.com/LianjiaTech/astra.

cs.CL↗

arXiv2026

ACDiT: Interpolating Autoregressive Conditional Modeling and Diffusion Transformer

Autoregressive and diffusion models have achieved remarkable progress in language models and visual generation, respectively. We present ACDiT, a novel Autoregressive blockwise Conditional Diffusion Transformer, that innovatively combines autoregressive and diffusion paradigms for continuous visual information. By introducing a block-wise autoregressive unit, ACDiT offers a flexible interpolation between token-wise autoregression and full-sequence diffusion, bypassing the limitations of discrete tokenization. The generation of each block is formulated as a conditional diffusion process, conditioned on prior blocks. ACDiT is easy to implement, as simple as applying a specially designed Skip-Causal Attention Mask on the standard diffusion transformer during training. During inference, the process iterates between diffusion denoising and autoregressive decoding that can make full use of KV-Cache. We validate the effectiveness of ACDiT on image, video, and text generation and show that ACDiT performs best among all autoregressive baselines under similar model scales on visual generation tasks. We also demonstrate that, benefiting from autoregressive modeling, pretrained ACDiT can be transferred in visual understanding tasks despite being trained with the generative objective. The analysis of the trade-off between autoregressive and diffusion demonstrates the potential of ACDiT to be used in long-horizon visual generation tasks. We hope that ACDiT offers a novel perspective on visual autoregressive generation and sheds light on new avenues for unified models.

cs.CV↗

arXiv2026

Gear-tuning meta-shaft for low-frequency torsional vibration suppression

Metastructures with band gaps provide a new solution for the torsional vibration attenuation in shaft systems, while tunable band gaps remain challenging, typically relying on additional physical fields or complex assembly processes. In this study, a gear-shifting tunable meta-shaft with self-locking gear (SLG) resonators is proposed, where a simple gear-shifting mechanism replaces complex tuning methods to achieve low-frequency torsional vibration suppression in tunable frequency ranges. First, as the key components of the SLG resonator, six inner curved beams are designed to provide precisely tunable torsional stiffness of the resonator through their deformed shapes controlled by shifting the gear teeth on the edge of the resonator. Also, based on the gear-shifting mechanism, the SLG resonator achieves resonant frequency modulation and opens tunable low-frequency torsional band gaps consistent with theoretical predictions. Then, the SLG resonators are periodically attached to a uniform shaft to construct a gear-shifting tunable meta-shaft, whose dynamic response is obtained using numerical simulations to evaluate its torsional wave attenuation performance. Finally, a gear-shifting tunable meta-shaft prototype is fabricated and experiments are carried out to study the propagation characteristics of torsional waves therein. Through the consistency observed among theoretical analysis, numerical simulations, and experimental results, the gear-shifting tunable meta-shaft is found to exhibit excellent attenuation performance in the tunable low-frequency band gaps. Therefore, the proposed gear-shifting tunable meta-shaft paves a new way for low-frequency torsional vibration suppression.

physics.app-ph↗

arXiv2026

Lip-Siri: Contactless Open-Sentence Silent Speech with Wi-Fi Backscatter

Silent speech interfaces (SSIs) enable silent interaction in noise-sensitive or privacy-sensitive settings. However, existing SSIs face practical deployment trade-offs among privacy, user experience, and energy consumption, and most remain limited to closed-set recognition over small, pre-defined vocabularies of words or sentences, which restricts real-world expressiveness. In this paper, we present Lip-Siri, to the best of our knowledge, the first Wi-Fi backscatter--based SSI that supports open-vocabulary sentence recognition via lexicon-guided subword decoding. Lip-Siri designs a frequency-shifted backscatter tag to isolate tag-modulated reflections and suppress interference from non-target motions, enabling reliable extraction of lip-motion traces from ubiquitous Wi-Fi signals. We then segment continuous traces into lip-motion units, cluster them, learn robust unit representations via cluster-based self-supervision, and finally propose a lexicon-guided Transformer encoder--decoder with beam search to decode variable-length sentence sequences. We implement an end-to-end prototype and evaluate it with 15 participants on 340 sentences and 3,398 words across multiple scenarios. Lip-Siri achieves 85.61% accuracy on word prediction and a WER of 36.87% on continuous sentence recognition, approaching the performance of representative vision-based lip-reading systems.

cs.HC↗

arXiv2026

Wavelet-Driven Masked Multiscale Reconstruction for PPG Foundation Models

Wearable foundation models have the potential to transform digital health by learning transferable representations from large-scale biosignals collected in everyday settings. While recent progress has been made in large-scale pretraining, most approaches overlook the spectral structure of photoplethysmography (PPG) signals, wherein physiological rhythms unfold across multiple frequency bands. Motivated by the insight that many downstream health-related tasks depend on multi-resolution features spanning fine-grained waveform morphology to global rhythmic dynamics, we introduce Masked Multiscale Reconstruction (MMR) for PPG representation learning - a self-supervised pretraining framework that explicitly learns from hierarchical time-frequency scales of PPG data. The pretraining task is designed to reconstruct randomly masked out coefficients obtained from a wavelet-based multiresolution decomposition of PPG signals, forcing the transformer encoder to integrate information across temporal and spectral scales. We pretrain our model with MMR using ~17 million unlabeled 10-second PPG segments from ~32,000 smartwatch users. On 17 of 19 diverse health-related tasks, MMR trained on large-scale wearable PPG data improves over or matches state-of-the-art open-source PPG foundation models, time-series foundation models, and other self-supervised baselines. Extensive analysis of our learned embeddings and systematic ablations underscores the value of wavelet-based representations, showing that they capture robust and physiologically-grounded features. Together, these results highlight the potential of MMR as a step toward generalizable PPG foundation models.

cs.LG↗

arXiv2026

Significant impact of Al1-xGaxN interlayer on GaN/AlN thermal boundary conductance

AlN-GaN heterostructures are central to high-power and high-frequency electronics, including RF devices, power converters, and AI accelerators. An intermediate Al1-xGaxN (AlGaN) layer is often present, either unintentionally during growth or intentionally to induce a 2D electron gas, yet its impact on the interfacial thermal boundary conductance (TBC) remains unknown due to the lack of reliable measurement or modeling methods. Here, we report a first principles-based evaluation of the TBCs of AlN-AlGaN, AlGaN-GaN, and AlN-AlGaN-GaN interfaces over the full alloy range. This is realized by the development of accurate deep learning interatomic potentials based on first-principles simulations. Contrary to other material systems where mixed interlayers enhance thermal coupling, we find that an AlGaN interlayer markedly degrades TBC between GaN and AlN, explaining the observation in experiments. Finally, we show that if the Al composition is sigmoidally transitioned from 0 to 1 across the AlN-GaN interface, it can remarkably increase the TBC, compared to an abrupt or a linear transition. This work is expected to shed light on an accurate thermal analysis and electro-thermal co-design of future AlGaN-based devices.

cond-mat.mtrl-sci↗