Search arXiv⌕ Search

arXiv subjects

Hao Zhou

Publications and source records attributed to Hao Zhou.

At least 73 records · Page 4Linked to original sources
arXiv2026

Diffusion Models for Smarter UAVs: Decision-Making and Modeling

Uncrewed Aerial Vehicles (UAVs) are increasingly used in modern communication networks. However, challenges in decision-making and digital modeling continue to hinder their rapid development. Reinforcement Learning (RL) algorithms face limitations such as low sample efficiency and limited data versatility, which are further amplified in UAV communications scenarios. Additionally, Digital Twin (DT) modeling presents significant challenges in decision-making and data management. RL models, often integrated into DT frameworks to address these issues, require large amounts of training data to make accurate predictions. Unlike traditional approaches that focus on class boundaries, Diffusion Models (DMs)-a new class of generative AI-learn the underlying probability distribution from training data and can generate reliable new patterns based on this learned distribution. DT and RL have complementary roles in enabling intelligent, data-driven UAV operations. DMs further enhance this synergy by addressing data scarcity, improving modeling accuracy, and generating realistic scenarios, which benefit both DT simulations and RL training. In this paper, we explore the integration of DMs with RL and DT. Simulation results confirm the effectiveness and benefits of DMs in generating neighbor velocity estimates in a four-UAV swarm coordination task using Deep Reinforcement Learning (DRL).

cs.LG↗
arXiv2026

LAPF: LLM-Agent-Based Path Finder Using the UAVScenes Dataset

Uncrewed aerial vehicles (UAVs) are increasingly deployed for autonomous navigation in complex outdoor environments, where dynamic conditions and mission requirements require intelligent adaptive decision-making. Existing optimization-based, Machine Learning (ML), and Reinforcement Learning (RL) approaches often rely on predefined models or task-specific training, limiting their generalization and adaptability in uncertain scenarios. Recent Large Language Model (LLM)-assisted approaches offer promising reasoning capabilities but remain constrained by limited agentic functionality, including insufficient memory, planning, and tool interaction mechanisms.This paper proposes an LLM-Agent-Based Path Finder (LAPF) framework for autonomous UAV navigation in town-scale outdoor environments. LAPF extends LLM-assisted navigation by integrating perception, memory, planning, and action modules into a closed-loop cognitive architecture. The proposed agent leverages prior navigation experiences, performs Chain-of-Thought (CoT) reasoning, couples each detected hazard to a bounded corrective action, and dynamically refines waypoint decisions based on environmental feedback.The three independent trials per method demonstrate that LAPF achieves mean path lengths of 512.83 m and 506.37 m, compared to the straight-line optimum of 497.33 m, corresponding to path length reductions of 17.2% and 15.6% relative to CoT prompting and absolute path efficiencies of 97.1% and 98.1% in open-field and obstacle-injected scenarios, respectively. Furthermore, LAPF is the only evaluated approach that couples every detected hazard to a bounded, metric-neutral corrective action while maintaining near-goal stability, with zero clamp events in both scenarios, whereas CoT prompting increases from 9.7 to 14.0 events.

cs.RO↗
arXiv2026

Envs-FORGE: Frontier-Optimized Reward-Grounded Environment Synthesis for Agent RL

Reinforcement learning (RL) for terminal agents needs executable training environments with reliable rewards and useful difficulty. Fixed recipes such as few-shot, Self-Instruct, and Evol-Instruct apply the same prompting policy to every seed, even when the current policy would benefit from a harder, easier, or simply different task. We present Envs-FORGE, a prompting policy that converts verifier rewards into per-seed environment-synthesis actions. Envs-FORGE estimates seed pass rates, scores six projection--direction actions around a target learning frontier, and solves a per-seed mixed-integer linear program (MILP) to choose the action that conditions generation. The selected action drives synchronized rewriting of the instruction, fixtures, oracle solution, tests, and Docker environment; only gold-verified bundles enter RL training. The indexed MILP form also supports optional soft skill coverage for portfolio planning. On Qwen 3.5 35B, Envs-FORGE improves Pass@1 over Base by 9.2 percentage points on tb-core (40.0% to 49.2%) and 6.4 points on tb-2.0 (23.0% to 29.4%), exceeding the strongest fixed-recipe baseline by 2.4 and 2.1 points. It reaches 77.1% on SWE-bench Verified versus 73.4% for Base, and improves tb-core by 6.8--9.2 points across the evaluated 4B--35B models. All synthesis methods export 100 verified environments and use 2.27M--2.88M synthesis tokens, placing the comparison at the same downstream training-set size and the same operational scale. The source code is available at https://github.com/DataArcTech/DataArc-SynData-Toolkit/.

cs.CL↗
arXiv2026

Evo-Bench: Can Language Models Improve Agent Harness?

Large Language Models (LLMs) have driven rapid progress in autonomous agents, yet standard evaluations remain confined to static task solving. An emerging frontier is harness evolution---the agent's capacity to autonomously optimize its own operating harness. However, systematically benchmarking this capability remains challenging, as existing evaluations fail to isolate harness improvements from base model strength, prevent task-specific overfitting, or capture long-horizon iterative research. To address these challenges, we introduce Evo-Bench, the first benchmark designed to evaluate models' intrinsic harness-evolving capabilities across Search, Office, and General agent domains. To rigorously isolate this capability, Evo-Bench employs a novel harness-guided construction framework: it leverages auxiliary-task evolution to identify tasks genuinely sensitive to framework improvements, followed by sensitivity-aware stratified splitting to ensure robust cross-suite generalization. Extensive evaluations across nine frontier and open-weight models reveal that top models achieve massive absolute gains reaching 16.6 points, closely approaching state-of-the-art human-engineered baselines. Crucially, while autonomous evolution outpeforms artificial harness in General tasks and excels in Search tasks, it struggles in Office tasks that demand highly specific processing workflows. Furthermore, our analysis exposes critical temporal anomalies like early saturation, while demonstrating that the synthesized harnesses act as highly transferable reasoning structures, consistently boosting diverse policy models.

cs.CL↗
arXiv2026

AIC-VDS: Attention-Based In-Context Learning for Joint Velocity Control and Data Collection Scheduling in Multi-UAV-Assisted Pipeline Monitoring

Uncrewed aerial vehicles (UAVs) are increasingly deployed for autonomous inspection and sensor data collection in large-scale infrastructure monitoring applications, such as pipeline monitoring, where timely anomaly detection is critical. Jointly optimizing data-collection schedules and flight velocities is a critical challenge, as inefficiencies can increase packet loss and inspection latency. While online deep reinforcement learning (DRL) is a widely investigated approach, it suffers from low sample efficiency, substantial training requirements, and simulation-to-reality gaps in time-sensitive scenarios. Large language models (LLMs) offer a promising alternative through in-context learning (ICL); however, their substantial input requirements can introduce considerable computational and communication overhead. To address this, we propose Attention-Based In-Context Learning for Velocity Control and Data Collection Scheduling (AIC-VDS), a joint optimization framework designed to minimize packet loss under partial and potentially outdated local network-state information. AIC-VDS utilizes an attention module to process real-time network-state data, including sensor battery levels, sensor queue lengths, communication channel conditions, UAV locations, time since the previous sensor visit, and sensor urgency scores. This module extracts task-relevant features to reduce input overhead before querying the LLM. The LLM leverages these compressed natural-language prompts to generate adaptive data-collection schedules and velocity-control decisions for UAV execution. Simulation results show that the attention-based representation reduces the average prompt length by 50\%, while AIC-VDS rapidly stabilizes packet loss in the considered scenario.

cs.AI↗
arXiv2026

JoVA: Unified Multimodal Learning for Joint Video-Audio Generation and Editing

In this paper, we present JoVA, a streamlined framework that unifies joint video-audio generation and editing. While existing methods often rely on fragmented, task-specific architectures or complex fusion mechanisms, JoVA employs native joint representation learning for direct video, audio, and text interaction in a dual-branch architecture. This design eliminates redundant alignment modules and effectively unifies diverse multimodal tasks within a single model. Furthermore, we utilize channel-wise conditioning for flexible image and video reference to avoid massive token expansion, alongside a mouth-area loss to enhance lip alignment. To fully empower and systematically evaluate this framework, we construct a comprehensive training corpus encompassing video-audio generation and editing datasets, and introduce unified benchmarks tailored for these multimodal tasks. Extensive experiments demonstrate that JoVA achieves state-of-the-art performance across benchmarks, establishing it as an extensible framework for versatile content creation. Project page: https://visual-ai.github.io/jova

cs.CV↗
arXiv2026

MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent

Despite improvements by length extrapolation, efficient attention and memory modules, handling infinitely long documents with linear complexity without performance degradation during extrapolation remains the ultimate challenge in long-text processing. We directly optimize for long-text tasks in an end-to-end fashion and introduce a novel agent workflow, MemAgent, which reads text in segments and updates the memory using an overwrite strategy. We extend the DAPO algorithm to facilitate training via independent-context multi-conversation generation. MemAgent has demonstrated superb long-context capabilities, being able to extrapolate from an 8K context trained on 32K text to a 3.5M QA task with performance loss < 5% and achieves 95%+ in 512K RULER test.

cs.CL↗
arXiv2026

Readability-Robust Code Summarization via Meta Curriculum Learning

Code summarization has emerged as a fundamental technique in the field of program comprehension. While code language models have shown significant advancements, the current models and benchmarks are confined to high-readability code, which contains sufficient semantic cues such as function and variable names. In the real world, however, code is often poorly structured or obfuscated, significantly degrading model performance. In this paper, we first empirically evaluate the robustness of state-of-the-art language models on poor-readability code for the task of code summarization, focusing on (1) their effectiveness, (2) the impact of prompt engineering, and (3) the robustness of different variants. Experimental results reveal that state-of-the-art models-including GPT-4o and DeepSeek-V3 experience a substantial performance drop when faced with poorly readable code, and that prompt engineering and reasoning-enhanced models offer limited improvements. Motivated by these findings, we propose RoFTCodeSum, a novel fine-tuning method that enhances the robustness of code summarization against poorly readable code. RoFTCodeSum marries the concepts of curriculum learning and meta-learning: based on the original dataset for fine-tuning, it creates curricular training sets, e.g., obfuscating function names and identifiers from the code, respectively, that have progressive difficulty in code comprehension. In each training step, the approach meta-updates the gradients using these progressively challenging datasets, thereby optimizing both accuracy and readability robustness simultaneously. Experimental results demonstrate that RoFTCodeSum exhibits increased robustness against semantic perturbation while enhancing performance on the original code.

cs.SE↗
arXiv2026

Stabilized Morley FEM for surface Stokes in stream-function formulation: Optimal convergence via a new geometric estimate

We propose an intrinsic stabilized Morley finite element method for the stream-function formulation of the surface Stokes problem on closed surfaces. The method is posed directly on a polyhedral approximation of the surface. A parameter-free jump stabilization is introduced to recover coercivity, and a discrete Korn's inequality for the trace-free Hessian is established. The main analytical contribution is a normal-separated geometric estimate for piecewise linearly approximated closed surfaces. It shows that a broad class of normal-dependent geometric consistency errors is in fact second order, even though a direct treatment suggests only first-order control. The missing order is recovered through an integral cancellation in the first normal variation. This estimate refines the standard $P_hν$-type estimate and yields second-order consistency for the trace-free Hessian, the Stokes-type tensor Green identity, and the conormal fluxes arising in the discretization. Together with the discrete Korn's inequality, these estimates yield, under the natural $H^3$ regularity, optimal-order convergence: first order in the broken $H^2$ norm and second order in the broken $H^1$ norm. As a direct consequence, the recovered tangential velocity admits a second-order $L^2$ estimate. Numerical experiments are provided to support the theoretical results.

math.NA↗
arXiv2026

BATON: A Multimodal Benchmark for Bidirectional Automation Transition Observation in Naturalistic Driving

Existing driving automation (DA) systems on production vehicles rely on human drivers to decide when to engage DA while requiring them to remain continuously attentive and ready to intervene. This design demands substantial situational judgment and imposes significant cognitive load, leading to steep learning curves, suboptimal user experience, and safety risks from both over-reliance and delayed takeover. Predicting when drivers hand over control to DA and when they take it back is therefore critical for designing proactive, context-aware HMI, yet existing datasets rarely capture the multimodal context, including road scene, driver state, vehicle dynamics, and route environment. To fill this gap, we introduce BATON, a large-scale naturalistic dataset capturing real-world DA usage across 127 drivers, and 136.6 hours of driving. The dataset synchronizes front-view video, in-cabin video, decoded CAN bus signals, radar-based lead-vehicle interaction, and GPS-derived route context, forming a closed-loop multimodal record around each control transition. We define three benchmark tasks: driving action understanding, handover prediction, and takeover prediction, and evaluate baselines spanning sequence models, classical classifiers, and zero-shot VLMs. Results show that visual input alone is insufficient for reliable transition prediction: front-view video captures road context but not driver state, while in-cabin video reflects driver readiness but not the external scene. Incorporating CAN and route-context signals substantially improves performance over video-only settings, indicating strong complementarity across modalities. We further find takeover events develop more gradually and benefit from longer prediction horizons, whereas handover events depend more on immediate contextual cues, revealing an asymmetry with direct implications for HMI design in assisted driving systems.

cs.HC↗
arXiv2026

Efficient Multilingual Reasoning Transfer via Progressive Code-Switching

Large reasoning models (LRMs) have achieved strong reasoning capabilities in English, yet their performance degrades significantly when required to reason in other languages. A natural solution is to transfer the model's English reasoning ability to target languages. However, existing transfer approaches typically rely on distilled target-language reasoning traces from stronger LRMs or online supervision from external judge models, which are costly and difficult to scale. In this paper, we propose PCS (Progressive Code-Switching), a more efficient transfer framework that requires only lightweight translation without any stronger model for distillation or judging. PCS first constructs code-switched reasoning traces by translating a subset of English reasoning steps into the target language, and uses them to initialize the model's code-switching ability via supervised fine-tuning. It then applies reinforcement learning with a step-level language consistency curriculum, progressively raising the target-language ratio until the model reasons entirely in the target language. This progressive design provides a smooth transfer path that avoids the instability and performance degradation commonly observed when directly enforcing target-language reasoning. Experiments on multiple benchmarks and five typologically diverse languages show that PCS substantially narrows the performance gap between target-language and English reasoning, yielding more language-consistent reasoning while maintaining competitive accuracy.

cs.CL↗
arXiv2026

Nanbeige4.2-3B: Unlocking Agentic Capabilities in a Compact Model

We present Nanbeige4.2-3B, a compact general agentic model with 3B non-embedding parameters. It delivers strong performance across code-agent, office-agent, and complex tool-use tasks while maintaining highly competitive reasoning capabilities in mathematics, coding, and science. Nanbeige4.2-3B is pretrained from scratch on 28T tokens with a Looped Transformer that reuses the layer stack to increase capacity without adding parameters. For SFT data and trajectory construction, we expand the diversity of executable environments, task assets, and agentic scaffolds through real-world deployment and large-scale synthesis. Our RL pipeline applies mixed-mode RLHF over Think and Non-Think responses to improve overall model quality and reduce failure cases, length-controlled reasoning RL to balance accuracy and reasoning efficiency, and agentic RL with outcome and process rewards to stabilize long-horizon training. Extensive evaluations show that Nanbeige4.2-3B outperforms larger models, including Qwen3.5-9B and Gemma4-12B, across diverse agentic benchmarks while remaining competitive on reasoning and alignment tasks. Performance with OpenClaw further supports its use as a compact local personal assistant.

cs.AI↗
arXiv2026

DriveDNA: A Large-Scale Multimodal Naturalistic Driving Dataset and Benchmark for Driving Style Identification

Driving style captures stable, driver-specific patterns in how a vehicle is driven. In naturalistic data, however, this signal is hard to isolate because drivers are observed in different vehicles, on different roads, and under different conditions, so models may mistake vehicle- or situation-specific regularities for driver-specific style. We introduce DriveDNA, a large-scale naturalistic dataset and benchmark for personalized driving-style modeling, comprising 4,121 drives from 465 drivers across 115 vehicle models and totaling 975 hours of human-controlled driving at 10 Hz with forward video, collected from community drivers in everyday use. DriveDNA defines driving style as a consistent, driver-specific behavioral pattern in how a vehicle moves under similar conditions. The benchmark evaluates this signal through three core tasks: few-shot driver re-identification, personalized behavior prediction, and condition-matched comparison, and provides behavioral annotations plus 276,248 rule-generated maneuver events across six classes with large-scale human auditing. We evaluate baselines spanning classical descriptors, supervised and self-supervised time-series encoders, multimodal fusion, probabilistic prediction, and zero-shot foundation models under a fixed multi-seed protocol. Learned representations substantially outperform classical descriptors on unseen drivers (AUROC .935 vs. .707) and retain driver-specific information under matched driving conditions, while descriptor performance approaches chance. Video-only models achieve comparable re-identification accuracy but exhibit severe route leakage, showing that strong recognition may arise from contextual shortcuts rather than driving behavior. These findings show that reliable driving-style evaluation must assess both the behavioral value of learned representations and their robustness to vehicle, drive, and condition confounds.

cs.LG↗
arXiv2026

Compiler-Grounded Hierarchical Diagnosis for LLM-Based Triton Kernel Optimization

Recent advances in large language models (LLMs) have enabled automated kernel generation and optimization, but most existing approaches rely on surface signals such as compilation feedback and profiling metrics. These signals reveal that a kernel is slow, but not why the backend compiler fails to realize a profitable optimization, especially on emerging accelerators such as NPUs. We therefore formulate kernel optimization as a progressive cross-layer diagnosis problem that links runtime symptoms to IR structure and compiler behavior before rewriting source. Based on this insight, we present our system, a compiler-grounded and hierarchical optimization framework for Triton kernels. the system escalates from lightweight pattern triage and profiling diagnosis to IR attribution and compiler-grounded analysis only when deeper evidence is needed, then proposes evidence-backed source-level rewrites. We implement the system on Triton for Ascend NPUs and evaluate it on 37 successfully converted entries from a standardized NPUKernelBench-derived Ascend 950 benchmark. Across these entries, the system attains a geometric-mean speedup of 4.35$\times$ and a median speedup of 2.73$\times$ from the initial to optimized Triton kernel; 22/37 exceed 2$\times$ and 13/37 exceed 5$\times$. The complete distribution ranges from near-baseline entries to large wins, motivating transparent reporting of the current system's scope and limitations.

cs.AI↗
arXiv2026

Intelligent Multi-UAV Navigation in ITNTNs: A Hierarchical LLM Approach

The deployment of high-speed Uncrewed Aerial Vehicles (UAVs) in 3D aerial highways necessitates robust coordination of physical flight kinematics and multi-tier network handovers. While Deep Reinforcement Learning (DRL) offers rapid tactical control, it lacks the zero-shot strategic reasoning required to quickly adapt to dynamic Integrated Terrestrial and Non-Terrestrial Networks (ITNTNs). Conversely, Large Language Models (LLMs) excel at semantic reasoning but suffer from high inference latency, rendering them unsuitable for real-time aerodynamic control. To bridge this gap, we propose a novel Hierarchical LLM-driven control framework. A massive cloud-based LLM deployed on a High-Altitude Platform Station (HAPS) manages slow-timescale global load balancing, while lightweight edge-LLMs on individual UAVs translate local observations into tactical sub-goals. These sub-goals guide a fast-timescale physical DRL controller to execute collision-free, handover-aware trajectories. Simulation results demonstrate that our agentic architecture significantly reduces collision rates and improves aggregate system throughput compared to existing baselines.

cs.RO↗
arXiv2026

Certifiable Safe Model-Based Reinforcement Learning with Control-Affine Dynamics Approximation

Safe model-based reinforcement learning (RL) often bridges control-theoretic analysis and RL for robots to safely explore (partially) unknown system dynamics while deriving control actions for task efficiency. The control performance and safety assurance typically rely on prior knowledge of partially modeled nominal system dynamics and the data-driven models that compensate for residual model uncertainties. However, existing methods often overlook the structure of residual model uncertainties (e.g., components affine in control), which could lead to overly conservative robot behaviors or invalid safety guarantees under the safe learning-based controllers. This paper proposes a safe reinforcement learning framework that learns control-affine dynamics with a certifiable data-driven safe policy using control barrier functions (CBF). Specifically, we first use Control-Affine Random Fourier Features (ARFF) to model robot dynamics in a control-affine form, which offers computational efficiency that scales with dataset size and reduces potential model bias for model-based reinforcement learning. Then, a model-free, efficient uncertainty quantification method using adaptive conformal prediction (ACP) is applied to quantify the uncertainty in the safety constraint arising from the learned control-affine dynamics. This allows for data-driven safety assurance amenable to principled and efficient controller synthesis with CBF. Simulation results on the cartpole and the 3D quadrotor platforms demonstrate the effectiveness of the proposed framework.

cs.RO↗
arXiv2026

Weak-to-Strong Generalization via Direct On-Policy Distillation

Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must generate many rollouts during training. As models scale, post-training itself becomes a bottleneck. We study a weak-to-strong alternative: run RL on a smaller model where rollouts are cheaper, then reuse what that RL run learned to improve a stronger target model. Directly distilling the post-RL weak teacher is not enough, because the teacher's final policy mixes useful RL gains with the limitations of the smaller model. We propose Direct On-Policy Distillation (Direct-OPD), which transfers the teacher's RL-induced policy shift instead. Direct-OPD compares the post-RL teacher with its own pre-RL reference and treats their log-ratio as a dense implicit reward for the student. In plain terms, the checkpoint pair tells us which actions RL made the weak model more or less likely to take, and Direct-OPD applies that signal on the stronger student's own on-policy states. This directly reuses the weak model's RL supervision signal without running sparse-reward RL on the target model. Empirically, Direct-OPD consistently leverages weaker teachers to improve stronger target models; notably, it boosts Qwen3-1.7B from 48.3% to 58.3% on AIME 2024 in just 4 hours on 8 A100 GPUs. It outperforms step-matched direct RL and enables the sequential composition of multiple policy shifts. Our results show that RL outcomes can be reused across model scales as implicit reward signals, not merely as final models to imitate.

cs.LG↗
arXiv2026

NKI-Agent: Domain-Specific Fine-Tuning and Agentic Tool Use for Neuron Kernel Generation

Recent agentic approaches to LLM-based kernel generation have achieved impressive results on CUDA. For emerging AI accelerators such as AWS Trainium and Inferentia, automated kernel generation and optimization remain largely unaddressed. Writing kernels for these chips via the Neuron Kernel Interface (NKI) is particularly challenging: developers must navigate a multi-engine architecture, tile-based programming, and explicit data movement across multi-level memory hierarchy. Moreover, no publicly-available training data, benchmarks, or tool-augmented agents exist for this domain. We introduce NKI-Agent, the first system combining domain-specific supervised fine-tuning (SFT) with a compile-verify-fix agent loop for NKI kernel generation. We adapt the existing CUDA-Agent framework to Neuron hardware, curate 6,000 NKI kernel generation tasks for training, and construct NKIBench, a 250-task benchmark across three difficulty levels. Evaluated on real Trn1 hardware, NKI-Agent with Claude Opus 4.8 and a rank-aware system prompt achieves a 77.3% pass rate on the 150-task NKIBench. We show that tool use is critical: Opus 4.8 scores 6% in single-shot mode without agent tools. On a 60-task subset, we show that an SFT-trained Qwen3-Coder-30B-A3B achieves 25.0% pass rate at 1/100th the cost, outperforming Claude Sonnet 4 (15.0%). We also report that Group Relative Policy Optimization (GRPO) with binary compilation reward fails to improve over SFT, providing guidance on reward design for RL-based kernel generation.

cs.LG↗