arXiv subjects

Zhiyuan Zhai

Publications and source records attributed to Zhiyuan Zhai.

16 recordsLinked to original sources

arXiv2026

Explainable AI for Next-Generation Wireless Physical Layer: Basics, State-of-the-Art, and Open Challenges

Next-generation wireless systems are expected to be ``AI-native," with neural networks (NNs) embedded throughout the physical (PHY) layer protocol stack to improve spectral efficiency, latency, and network autonomy. However, the opacity of deep learning (DL) models raises increasing concerns about system reliability, safety, and privacy, especially under complex and time-varying network environments. This survey studies explainable AI (XAI) in wireless PHY layers from the explainability perspective. We first formalize a series of responsibility-oriented goals for wireless XAI. Then, we develop a systematic taxonomy of explainability approaches and distill practical criteria for deploying explanations in communication scenarios. We provide a comprehensive review of where and how XAI can be applied throughout the PHY layer, connecting representative learning paradigms to appropriate explanation techniques, evaluation metrics, and deployment considerations. Open challenges and future directions are discussed, including explainability-performance tradeoffs, explainability-aware data processing, customized XAI for communication-specific structures, cross-layer explanation consistency, and emerging needs for explaining LLM- and Agentic-AI-driven PHY layers.

eess.SP↗

arXiv2026

Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL

Group-relative RL training (GRPO) samples a small group of parallel rollouts for every training prompt and uses their within-group reward spread to compute per-trajectory advantages. In agentic environments each rollout is a long multi-turn dialogue with one LLM call per step, so this multi-sample multiplier dominates the total training cost. When every rollout of a prompt ends with the same reward, the group has zero reward variance and contributes no gradient, so the extra rollouts add no information; such groups are common in practice (typically around 40% of all groups), so the wasted-compute fraction is substantial rather than marginal. Existing methods filter such groups at the prompt level, either after their rollouts are paid for or before any rollout begins, but both decide without using information that becomes available during the rollout itself. We instead ask whether the in-group divergence between the partial trajectories at an intermediate step can already predict that the group will be zero-variance: when the parallel rollouts have already converged on the same action prefix, the group is on track to produce a single reward, and we can stop early. We propose a one-parameter gate that stops a group when the mean pairwise prefix edit distance between its partial action sequences falls below a threshold. On a 60-iteration on-policy GRPO run on ALFWorld with Qwen2.5-7B, averaged over four random seeds, the gated arm finishes 10.7% faster in wall-clock (bootstrap 95% CI excludes 0) and shifts held-out success rate on 50 unseen tasks by +2.5 pp, with the held-out gain tracing to a measurable reduction in zero-advantage gradient-batch dilution. Code is available at https://github.com/zhiyuanZhai20/selective-rollout.

cs.LG↗

arXiv2026

Revisable by Design: A Theory of Streaming LLM Agent Execution

Current LLM agents operate under an implicit but universal assumption: execution is a transaction -- the user submits a request, the agent works in isolation, and only upon completion does the dialogue resume. This forces users into a binary choice: wait for a potentially incorrect output, or interrupt and lose all progress. We reject this assumption and propose the stream paradigm, in which agent execution and user intervention are concurrent, interleaved processes sharing a bidirectional channel. We formalize this paradigm through a reversibility taxonomy that classifies every agent action as Idempotent, Reversible, Compensable, or Irreversible, and arrive at a core conclusion: an agent's flexibility is bounded by its reversibility. We prove that conflicting compensable actions impose unavoidable adaptation costs and that conflicting irreversible actions make full specification satisfaction impossible -- these costs are properties of the action space, not of the algorithm. Guided by this insight, we present the Revision Absorber, a reactive algorithm based on the Earliest-Conflict Rollback rule that is structurally optimal under mild assumptions. Experiments on StreamBench with real LLM agents validate all predictions: the Absorber matches the quality of a brute-force full-restart baseline while wasting an order of magnitude fewer steps of already-completed work, turning mid-execution revisions from a dead-end into a first-class interaction.

cs.LG↗

arXiv2026

How Much Thinking is Enough? Quantifying and Understanding Redundancy in LLM Reasoning

Reasoning-capable large language models solve hard problems by emitting long chains of thought, paying heavily in latency, GPU time, and energy. Casual inspection of their traces reveals extensive reformulation, verification, and circular self-reflection, yet how much of this deliberation is actually necessary has never been measured at scale or explained from first principles. This paper closes both gaps. We formalise reasoning redundancy directly in terms of the reasoning model itself: the redundancy of a correct trace is the largest fraction of its trailing segmented steps that can be truncated while $π$, forced to terminate thinking and emit a final answer, still produces the correct answer. A large-scale quantification across four frontier reasoning models and two mathematical benchmarks shows that step-level redundancy is consistently high -- between 61% and 93% across the 8 (model, benchmark) conditions we study, with the median critical prefix equal to a single segmented step in six of the eight conditions -- that the finding is robust to the choice of judge family, and that although $ρ$ decreases with problem difficulty on MATH-500, all four models remain substantially redundant ($ρ\in [46\%, 85\%]$) even on the hardest Level-5 problems. We then prove that this redundancy is a structural consequence of length-agnostic outcome rewards, not a model-specific artefact: under any such reward, no finite expected stopping time is optimal. The result holds regardless of RL algorithm, base model, data distribution, or whether the policy is obtained via RL or distillation; over-thinking is therefore not a bug to be patched in individual models but a structural property of how current reasoning models are trained. Code: https://github.com/zhiyuanZhai20/how-much-thinking-is-enough

cs.AI↗

arXiv2026

Adaptive Test-Time Compute Allocation for Reasoning LLMs via Constrained Policy Optimization

Test-time compute scaling, the practice of spending extra computation during inference via repeated sampling, search, or extended reasoning, has become a powerful lever for improving large language model performance. Yet deploying these techniques under finite inference budgets requires a decision that current systems largely ignore: which inputs deserve more compute, and which can be answered cheaply? We formalize this as a constrained optimization problem (maximize expected accuracy subject to an average compute budget) and solve it with a two-stage Solve-then-Learn pipeline. In the solve stage, Lagrangian relaxation decomposes the global constraint into per-instance sub-problems, each admitting a closed-form oracle action that optimally prices accuracy against cost. We prove that the induced cost is monotone in the dual variable, enabling exact budget targeting via binary search. In the learn stage, a lightweight classifier is trained to predict oracle actions from cheap input features, amortizing the allocation rule for real-time deployment. We establish that the task-level regret of the learned policy is bounded by its imitation error times the worst-case per-instance gap, yielding a clean reduction from constrained inference to supervised classification. Experiments on MATH and GSM8K with three LLMs (DeepSeek-V3, GPT-4o-mini, Qwen2.5-7B) show that our method consistently outperforms uniform and heuristic allocation baselines, achieving up to 12.8% relative accuracy improvement on MATH under matched budget constraints, while closely tracking the Lagrangian oracle upper bound with over 91% imitation accuracy.

cs.LG↗

arXiv2026

Does RL Expand the Capability Boundary of LLM Agents? A PASS@(k,T) Analysis

Does reinforcement learning genuinely expand what LLM agents can do, or merely make them more reliable? For static reasoning, recent work answers the second: base and RL pass@k curves converge at large k. We ask whether this holds for agentic tool use, where T rounds of interaction enable compositional strategies that re-sampling cannot recover. We introduce PASS@(k,T), a two-dimensional metric that jointly varies sampling budget k and interaction depth T, separating capability expansion from efficiency improvement. Our main finding is that, contrary to the static-reasoning result, tool-use RL genuinely enlarges the capability boundary: the RL agent's pass-curve pulls above the base model's and the gap widens at large k rather than converging. The expansion is specific to compositional, sequential information gathering; on simpler tasks RL behaves as prior work predicts. Under matched training data, supervised fine-tuning regresses the boundary on the same compositional tasks, isolating self-directed exploration as the causal factor. Mechanism analysis shows RL reweights the base strategy distribution toward the subset whose downstream reasoning more often yields a correct answer, with the improvement concentrated on how the agent integrates retrieved information. These results reconcile optimistic and pessimistic readings of RL for LLMs: both are correct, on different task types.

cs.LG↗

arXiv2026

Evidential Quantum Vertical Federated Learning

Quantum federated learning (QFL) has recently emerged as a promising paradigm for privacy-preserving collaborative learning, yet most existing studies focus on horizontal federated learning and ignore the vertical federated learning (VFL), where parties hold complementary features of aligned samples. In this work, we propose Evidential Quantum Vertical Federated Learning (eviQVFL), a VFL-tailored QFL framework that employs a hybrid classical-quantum architecture for party-side feature processing, mapping local features into a quantum state. To preserve privacy and avoid information loss, party-side output states are directly transmitted to the server via quantum teleportation, and the server fuses the received quantum states with a non-parametric evidential fusion circuit grounded in evidence theory, followed by measurement-based inference. Extensive simulations on image classification and other real-world datasets demonstrate that eviQVFL consistently achieves higher classification accuracy than other classical and quantum baselines under comparable parameter budgets. Both empirical observations and theoretical analysis indicate that eviQVFL achieve less approximation error with limited quantum resources, while maintaining training stability and offering stronger feature privacy.

quant-ph↗

arXiv2026

Event-Triggered Gossip for Distributed Learning

While distributed learning offers a new learning paradigm for distributed network with no central coordination, it is constrained by communication bottleneck between nodes. We develop a new event-triggered gossip framework for distributed learning to reduce inter-node communication overhead. The framework introduces an adaptive communication control mechanism that enables each node to autonomously decide in a fully decentralized fashion when to exchange model information with its neighbors based on local model deviations. We analyze the ergodic convergence of the proposed framework under noconvex objectives and interpret the convergence guarantees under different triggering conditions. Simulation results show that the proposed framework achieves substantially lower communication overhead than the state-of-the-art distributed learning methods, reducing cumulative point-to-point transmissions by \textbf{71.61\%} with only a marginal performance loss, compared with the conventional full-communication baseline.

eess.SP↗

arXiv2025

Learning Performance Optimization for Edge AI System with Time and Energy Constraints

Edge AI, which brings artificial intelligence to the edge of the network for real-time processing and decision-making, has emerged as a transformative technology across various applications. However, the deployment of Edge AI systems faces significant challenges due to high energy consumption and extended operation time. In this paper, we consider an Edge AI system which integrates the data acquisition, computation and communication processes, and focus on improving learning performance of this system. We model the time and energy consumption of different processes and perform a rigorous convergence analysis to quantify the impact of key system parameters, such as the amount of collected data and the number of training rounds, on the learning performance. Based on this analysis, we formulate a system-wide optimization problem that seeks to maximize learning performance under given time and energy constraints. We explore both homogeneous and heterogeneous device scenarios, developing low-complexity algorithms based on one-dimensional search and alternating optimization to jointly optimize data collection time and training rounds. Simulation results validate the accuracy of our convergence analysis and demonstrate the effectiveness of the proposed algorithms, providing valuable insights into designing energy-efficient Edge AI systems under real-world conditions.

eess.SP↗

arXiv2025

Trajectory Design for UAV-Assisted Logistics Collection in Low-Altitude Economy

Low-altitude economy (LAE) is rapidly emerging as a key driver of innovation, encompassing economic activities taking place in airspace below 500 meters. Unmanned aerial vehicles (UAVs) provide valuable tools for logistics collection within LAE systems, offering the ability to navigate through complex environments, avoid obstacles, and improve operational efficiency. However, logistics collection tasks involve UAVs flying through complex three-dimensional (3D) environments while avoiding obstacles, where traditional UAV trajectory design methods,typically developed under free-space conditions without explicitly accounting for obstacles, are not applicable. This paper presents, we propose a novel algorithm that combines the Lin-Kernighan-Helsgaun (LKH) and Deep Deterministic Policy Gradient (DDPG) methods to minimize the total collection time. Specifically, the LKH algorithm determines the optimal order of item collection, while the DDPG algorithm designs the flight trajectory between collection points. Simulations demonstrate that the proposed LKH-DDPG algorithm significantly reduces collection time by approximately 49 percent compared to baseline approaches, thereby highlighting its effectiveness in optimizing UAV trajectories and enhancing operational efficiency for logistics collection tasks in the LAE paradigm.

eess.SP↗

arXiv2025

Integrated Sensing and Communication with UAV Swarms via Decentralized Consensus ADMM

UAV swarms can form virtual antenna arrays to exploit additional spatial degrees of freedom and enhance integrated sensing and communication (ISAC). The optimization of UAV positions is challenging due to the distributed nature of swarms and the lack of a global view at individual UAVs. This paper presents a new decentralized optimization framework that allows UAVs to decide their locations in parallel and reach consensus on a globally optimal swarm geometry for ISAC. Specifically, we derive the achievable uplink rate and Cramér-Rao Bound (CRB) as tractable metrics for communication and sensing, respectively. The UAV positions are optimized to balance maximizing the communication rate and minimizing the CRB. To solve this non-convex problem with coupled variables, we develop a decentralized consensus alternating direction method of multipliers (ADMM) algorithm, which enables the UAVs to iteratively align their local updates and reach consensus. The algorithm decomposes the global objective into local projection updates, proxy-assisted consensus coordination, and lightweight dual updates, ensuring scalability and consistency throughout the swarm. Simulations demonstrate that the proposed consensus ADMM algorithm converges rapidly with strong scalability, and that the UAV swarm significantly outperforms fixed-array baselines in both communication and sensing performance.

eess.SP↗

arXiv2025

Decentralized Federated Learning with Distributed Aggregation Weight Optimization

Decentralized federated learning (DFL) is an emerging paradigm to enable edge devices collaboratively training a learning model using a device-to-device (D2D) communication manner without the coordination of a parameter server (PS). Aggregation weights, also known as mixing weights, are crucial in DFL process, and impact the learning efficiency and accuracy. Conventional design relies on a so-called central entity to collect all local information and conduct system optimization to obtain appropriate weights. In this paper, we develop a distributed aggregation weight optimization algorithm to align with the decentralized nature of DFL. We analyze convergence by quantitatively capturing the impact of the aggregation weights over decentralized communication networks. Based on the analysis, we then formulate a learning performance optimization problem by designing the aggregation weights to minimize the derived convergence bound. The optimization problem is further transformed as an eigenvalue optimization problem and solved by our proposed subgradient-based algorithm in a distributed fashion. In our algorithm, edge devices only need local information to obtain the optimal aggregation weights through local (D2D) communications, just like the learning itself. Therefore, the optimization, communication, and learning process can be all conducted in a distributed fashion, which leads to a genuinely distributed DFL system. Numerical results demonstrate the superiority of the proposed algorithm in practical DFL deployment.

eess.SP↗

arXiv2025

Spectral-Convergent Decentralized Machine Learning: Theory and Application in Space Networks

Decentralized machine learning (DML) supports collaborative training in large-scale networks with no central server. It is sensitive to the quality and reliability of inter-device communications that result in time-varying and stochastic topologies. This paper studies the impact of unreliable communication on the convergence of DML and establishes a direct connection between the spectral properties of the mixing process and the global performance. We provide rigorous convergence guarantees under random topologies and derive bounds that characterize the impact of the expected mixing matrix's spectral properties on learning. We formulate a spectral optimization problem that minimizes the spectral radius of the expected second-order mixing matrix to enhance the convergence rate under probabilistic link failures. To solve this non-smooth spectral problem in a fully decentralized manner, we design an efficient subgradient-based algorithm that integrates Chebyshev-accelerated eigenvector estimation with local update and aggregation weight adjustment, while ensuring symmetry and stochasticity constraints without central coordination. Experiments on a realistic low Earth orbit (LEO) satellite constellation with time-varying inter-satellite link models and real-world remote sensing data demonstrate the feasibility and effectiveness of the proposed method. The method significantly improves classification accuracy and convergence efficiency compared to existing baselines, validating its applicability in satellite and other decentralized systems.

eess.SP↗

arXiv2024

UAV-Enabled Asynchronous Federated Learning

To exploit unprecedented data generation in mobile edge networks, federated learning (FL) has emerged as a promising alternative to the conventional centralized machine learning (ML). However, there are some critical challenges for FL deployment. One major challenge called straggler issue severely limits FL's coverage where the device with the weakest channel condition becomes the bottleneck of the model aggregation performance. Besides, the huge uplink communication overhead compromises the effectiveness of FL, which is particularly pronounced in large-scale systems. To address the straggler issue, we propose the integration of an unmanned aerial vehicle (UAV) as the parameter server (UAV-PS) to coordinate the FL implementation. We further employ over-the-air computation technique that leverages the superposition property of wireless channels for efficient uplink communication. Specifically, in this paper, we develop a novel UAV-enabled over-the-air asynchronous FL (UAV-AFL) framework which supports the UAV-PS in updating the model continuously to enhance the learning performance. Moreover, we conduct a convergence analysis to quantitatively capture the impact of model asynchrony, device selection and communication errors on the UAV-AFL learning performance. Based on this, a unified communication-learning problem is formulated to maximize asymptotical learning performance by optimizing the UAV-PS trajectory, device selection and over-the-air transceiver design. Simulation results demonstrate that the proposed scheme achieves substantially learning efficiency improvement compared with the state-of-the-art approaches.

eess.SP↗

arXiv2023

Decentralized Federated Learning via MIMO Over-the-Air Computation: Consensus Analysis and Performance Optimization

Decentralized federated learning (DFL), inherited from distributed optimization, is an emerging paradigm to leverage the explosively growing data from wireless devices in a fully distributed manner.DFL enables joint training of machine learning model under device to device (D2D) communication fashion without the coordination of a parameter server. However, the deployment of wireless DFL is facing some pivotal challenges. Communication is a critical bottleneck due to the required extensive message exchange between neighbor devices to share the learned model. Besides, consensus becomes increasingly difficult as the number of devices grows because there is no available central server to perform coordination. To overcome these difficulties, this paper proposes employing over-the-air computation (Aircomp) to improve communication efficiency by exploiting the superposition property of analog waveform in multi-access channels, and introduce the mixing matrix mechanism to promote consensus using the spectral property of symmetric doubly stochastic matrix. Specifically, we develop a novel multiple-input multiple-output over-the-air DFL (MIMO OA-DFL) framework to study over-the-air DFL problem over MIMO multiple access channels. We conduct a general convergence analysis to quantitatively capture the influence of aggregation weight and communication error on the MIMO OA-DFL performance in \emph{ad hoc} networks. The result shows that the communication error together with the spectral gap of mixing matrix has a significant impact on the learning performance. Based on this, a joint communication-learning optimization problem is formulated to optimize transceiver beamformers and mixing matrix. Extensive numerical experiments are performed to reveal the characteristics of different topologies and demonstrate the substantial learning performance enhancement of our proposed algorithm.

eess.SP↗

arXiv2022

Energy-Efficient UAV-Mounted RIS Assisted Mobile Edge Computing

Unmanned aerial vehicle (UAV) and reconfigurable intelligent surface (RIS) have been recently applied in the field of mobile edge computing (MEC) to improve the data exchange environment by proactively changing the wireless channels through maneuverable location deployment and intelligent signals reflection, respectively. Nevertheless, they may suffer from inherent limitations in practical scenarios. UAV-mounted RIS (U-RIS), as a promising integrated approach, can combine the advantages of UAV and RIS to break the limit. Inspired by this, we consider a novel U-RIS assisted MEC system, where a U-RIS is deployed to assist the communication between the ground users and an MEC server. The joint UAV trajectory, RIS passive beamforming and MEC resource allocation design is developed to maximize the energy efficiency (EE) of the system. To tackle the intractable non-convex problem, we divide it into two subproblems and solve them iteratively based on successive convex approximation (SCA) and the Dinkelbach method. Finally we obtain a high-performance suboptimal solution. Simulation results show that the proposed algorithm significantly improves the energy efficiency of the MEC system.

eess.SP↗