Zeyang Li

Zeyang Li, Chuxiong Hu, Weiye Zhao et al.

Ensuring safety of nonlinear systems under model uncertainty and external disturbances is crucial, especially for real-world control tasks. Predictive methods such as robust model predictive control (RMPC) require solving nonconvex optimization problems online, which leads to high computational burden and poor scalability. Reinforcement learning (RL) works well with complex systems, but pays the price of losing rigorous safety guarantee. This paper presents a theoretical framework that bridges the advantages of both RMPC and RL to synthesize safety filters for nonlinear systems with state- and action-dependent uncertainty. We decompose the robust invariant set (RIS) into two parts: a target set that aligns with terminal region design of RMPC, and a reach-avoid set that accounts for the rest of RIS. We propose a policy iteration approach for robust reach-avoid problems and establish its monotone convergence. This method sets the stage for an adversarial actor-critic deep RL algorithm, which simultaneously synthesizes a reach-avoid policy network, a disturbance policy network, and a reach-avoid value network. The learned reach-avoid policy network is utilized to generate nominal trajectories for online verification, which filters potentially unsafe actions that may drive the system into unsafe regions when worst-case disturbances are applied. We formulate a second-order cone programming (SOCP) approach for online verification using system level synthesis, which optimizes for the worst-case reach-avoid value of any possible trajectories. The proposed safety filter requires much lower computational complexity than RMPC and still enjoys persistent robust safety guarantee. The effectiveness of our method is illustrated through a numerical example.

Ranxu zhang, zeyang li, Jiacheng Huang et al.

Agentic reinforcement learning (Agentic RL) has achieved strong progress in tasks with clear success signals. However, many real-world agent applications require user-conditioned behavior: the same query may call for different planning strategies and tool-use decisions across users. This setting raises key challenges: generic rewards cannot capture heterogeneous user preferences, observed behaviors are entangled with conformity effects, and flat memories cannot support personalized skill retrieval. To this end, we propose a unified personalized Agentic RL framework that embeds personalization into training-time optimization. At its core is \emph{Personalized Anchor Reward-Decoupled Policy Optimization} (\textbf{PARPO}), which decouples generic task-quality rewards from personalized preference rewards and uses user-specific anchors to stabilize learning under heterogeneous reward scales. We further introduce a two-stage preference-disentangled reward model and \emph{Preference-Aligned Skill Evolution Graph Memory} (\textbf{PSGM}) for personalized supervision and preference-aligned skill retrieval. Together, they form a closed loop of preference identification, policy optimization, and structured skill accumulation. Experiments on ETAPP, ETAPP-Hard, and SJAgent show that our framework consistently outperforms strong memory and RL baselines. Code and data are included in the supplementary materials.

Yangang Ren, Yao Lyu, Wenxuan Wang et al.

Zero-sum Markov Games (MGs) has been an efficient framework for multi-agent systems and robust control, wherein a minimax problem is constructed to solve the equilibrium policies. At present, this formulation is well studied under tabular settings wherein the maximum operator is primarily and exactly solved to calculate the worst-case value function. However, it is non-trivial to extend such methods to handle complex tasks, as finding the maximum over large-scale action spaces is usually cumbersome. In this paper, we propose the smoothing policy iteration (SPI) algorithm to solve the zero-sum MGs approximately, where the maximum operator is replaced by the weighted LogSumExp (WLSE) function to obtain the nearly optimal equilibrium policies. Specially, the adversarial policy is served as the weight function to enable an efficient sampling over action spaces.We also prove the convergence of SPI and analyze its approximation error in $\infty -$norm based on the contraction mapping theorem. Besides, we propose a model-based algorithm called Smooth adversarial Actor-critic (SaAC) by extending SPI with the function approximations. The target value related to WLSE function is evaluated by the sampled trajectories and then mean square error is constructed to optimize the value function, and the gradient-ascent-descent methods are adopted to optimize the protagonist and adversarial policies jointly. In addition, we incorporate the reparameterization technique in model-based gradient back-propagation to prevent the gradient vanishing due to sampling from the stochastic policies. We verify our algorithm in both tabular and function approximation settings. Results show that SPI can approximate the worst-case value function with a high accuracy and SaAC can stabilize the training process and improve the adversarial robustness in a large margin.

Zeyang Li, Chuxiong Hu, Yunan Wang et al.

Reinforcement learning (RL) agents are vulnerable to adversarial disturbances, which can deteriorate task performance or compromise safety specifications. Existing methods either address safety requirements under the assumption of no adversary (e.g., safe RL) or only focus on robustness against performance adversaries (e.g., robust RL). Learning one policy that is both safe and robust remains a challenging open problem. The difficulty is how to tackle two intertwined aspects in the worst cases: feasibility and optimality. Optimality is only valid inside a feasible region, while identification of maximal feasible region must rely on learning the optimal policy. To address this issue, we propose a systematic framework to unify safe RL and robust RL, including problem formulation, iteration scheme, convergence analysis and practical algorithm design. This unification is built upon constrained two-player zero-sum Markov games. A dual policy iteration scheme is proposed, which simultaneously optimizes a task policy and a safety policy. The convergence of this iteration scheme is proved. Furthermore, we design a deep RL algorithm for practical implementation, called dually robust actor-critic (DRAC). The evaluations with safety-critical benchmarks demonstrate that DRAC achieves high performance and persistent safety under all scenarios (no adversary, safety adversary, performance adversary), outperforming all baselines significantly.

Zeyang Li, Sunbochen Tang, Navid Azizan

Diffusion and flow policies are gaining prominence in online reinforcement learning (RL) due to their expressive power, yet training them efficiently remains a critical challenge. A fundamental difficulty in online RL is the lack of direct samples from the target distribution; instead, the target is an unnormalized Boltzmann distribution defined by the Q-function. To address this, two seemingly distinct families of methods have been proposed for diffusion policies: a noise-expectation family, which utilizes a weighted average of noise as the training target, and a gradient-expectation family, which employs a weighted average of Q-function gradients. Yet, it remains unclear how these objectives relate formally or if they can be synthesized into a more general formulation. In this paper, we propose a unified framework, reverse flow matching (RFM), which rigorously addresses the problem of training diffusion and flow models without direct target samples. By adopting a reverse inferential perspective, we formulate the training target as a posterior mean estimation problem given an intermediate noisy sample. Crucially, we introduce Langevin Stein operators to construct zero-mean control variates, deriving a general class of estimators that effectively reduce importance sampling variance. We show that existing noise-expectation and gradient-expectation methods are two specific instances within this broader class. This unified view yields two key advancements: it extends the capability of targeting Boltzmann distributions from diffusion to flow policies, and enables the principled combination of Q-value and Q-gradient information to derive an optimal, minimum-variance estimator, thereby improving training efficiency and stability. We instantiate RFM to train a flow policy in online RL, and demonstrate improved performance on continuous-control benchmarks compared to diffusion policy baselines.

Zeyang Li, Kaveh Alim, Navid Azizan

Diffusion and flow-matching have emerged as powerful methodologies for generative modeling, with remarkable success in capturing complex data distributions and enabling flexible guidance at inference time. Many downstream applications, however, demand enforcing hard constraints on generated samples (for example, robot trajectories must avoid obstacles), a requirement that goes beyond simple guidance. Prevailing projection-based approaches constrain the entire sampling path to the constraint manifold, which is overly restrictive and degrades sample quality. In this paper, we introduce a novel framework that reformulates hard-constrained sampling as a trajectory optimization problem. Our key insight is to leverage numerical optimal control to steer the sampling trajectory so that constraints are satisfied precisely at the terminal time. By exploiting the underlying structure of flow-matching models and adopting techniques from model predictive control, we transform this otherwise complex constrained optimization problem into a tractable surrogate that can be solved efficiently and effectively. Furthermore, this trajectory optimization perspective offers significant flexibility beyond mere constraint satisfaction, allowing for the inclusion of integral costs to minimize distribution shift and terminal objectives to further enhance sample quality, all within a unified framework. We provide a control-theoretic analysis of our method, establishing bounds on the approximation error between our tractable surrogate and the ideal formulation. Extensive experiments across diverse domains, including robotics (planning), partial differential equations (boundary control), and vision (text-guided image editing), demonstrate that our algorithm, which we name $\textit{HardFlow}$, substantially outperforms existing methods in both constraint satisfaction and sample quality.

Zeyang Li, Chuxiong Hu, Shengbo Eben Li et al.

Safety is a primary concern when applying reinforcement learning to real-world control tasks, especially in the presence of external disturbances. However, existing safe reinforcement learning algorithms rarely account for external disturbances, limiting their applicability and robustness in practice. To address this challenge, this paper proposes a robust safe reinforcement learning framework that tackles worst-case disturbances. First, this paper presents a policy iteration scheme to solve for the robust invariant set, i.e., a subset of the safe set, where persistent safety is only possible for states within. The key idea is to establish a two-player zero-sum game by leveraging the safety value function in Hamilton-Jacobi reachability analysis, in which the protagonist (i.e., control inputs) aims to maintain safety and the adversary (i.e., external disturbances) tries to break down safety. This paper proves that the proposed policy iteration algorithm converges monotonically to the maximal robust invariant set. Second, this paper integrates the proposed policy iteration scheme into a constrained reinforcement learning algorithm that simultaneously synthesizes the robust invariant set and uses it for constrained policy optimization. This algorithm tackles both optimality and safety, i.e., learning a policy that attains high rewards while maintaining safety under worst-case disturbances. Experiments on classic control tasks show that the proposed method achieves zero constraint violation with learned worst-case adversarial disturbances, while other baseline algorithms violate the safety constraints substantially. Our proposed method also attains comparable performance as the baselines even in the absence of the adversary.

Yunan Wang, Jizhou Yan, Chuxiong Hu et al.

Optimal path parameterization (OPP) is a fundamental problem for planning trajectories along a prescribed geometric path under kinodynamic constraints and task-dependent objectives. While TOPP minimizes traversal time, its saturating states and controls may induce vibration and tracking errors, which can be mitigated by introducing smoothness objectives. However, a key capability gap remains in OPP: feasibility guarantees, general-objective optimality certificates, and computational efficiency are difficult to achieve simultaneously in a unified framework, especially for third-order OPP (OPP3) with non-convex constraints. This paper proposes reachability-augmented dual dynamic programming (RDDP), a state-grid-free and objective-aware DP framework for OPP. The key idea is to replace the relatively complete recourse assumption used in classical dual DP (DDP) with OPP-specific backward reachable sets, and then generate both value-function cuts and trial trajectories only inside these reachable sets. For convex and non-convex OPP, we prove global optimality and Karush-Kuhn-Tucker convergence of RDDP under OPP-specific conditions, respectively. Efficient instantiations are developed for OPP2 and OPP3. Experiments show that RDDP achieves objective values comparable to convex-optimization baselines while reducing computation time by 28.6 times for OPP2 and 5.8 times for OPP3. RDDP also achieves faster convergence than grid-based DP. Compared with reachability-analysis methods, RDDP retains the reachability mechanism while replacing local maximum-control propagation with value-function-guided control selection, thereby enabling objectives beyond traversal time. In summary, RDDP addresses a key capability gap in OPP by unifying certifiable general-objective optimization, reachability-based feasibility preservation, and online-compatible low-dimensional DP computation in a single OPP framework.

Shuaimin Li, Liyang Fan, Yufang Lin et al.

Existing paper review methods often rely on superficial manuscript features or directly on large language models (LLMs), which are prone to hallucinations, biased scoring, and limited reasoning capabilities. Moreover, these methods often fail to capture the complex argumentative reasoning and negotiation dynamics inherent in reviewer-author interactions. To address these limitations, we propose ReViewGraph (Reviewer-Author Debates Graph Reasoner), a novel framework that performs heterogeneous graph reasoning over LLM-simulated multi-round reviewer-author debates. In our approach, reviewer-author exchanges are simulated through LLM-based multi-agent collaboration. Diverse opinion relations (e.g., acceptance, rejection, clarification, and compromise) are then explicitly extracted and encoded as typed edges within a heterogeneous interaction graph. By applying graph neural networks to reason over these structured debate graphs, ReViewGraph captures fine-grained argumentative dynamics and enables more informed review decisions. Extensive experiments on three datasets demonstrate that ReViewGraph outperforms strong baselines with an average relative improvement of 15.73%, underscoring the value of modeling detailed reviewer-author debate structures.

Zeyang Li, Chuxiong Hu, Yunan Wang et al.

Regularization is one of the most important techniques in reinforcement learning algorithms. The well-known soft actor-critic algorithm is a special case of regularized policy iteration where the regularizer is chosen as Shannon entropy. Despite some empirical success of regularized policy iteration, its theoretical underpinnings remain unclear. This paper proves that regularized policy iteration is strictly equivalent to the standard Newton-Raphson method in the condition of smoothing out Bellman equation with strongly convex functions. This equivalence lays the foundation of a unified analysis for both global and local convergence behaviors of regularized policy iteration. We prove that regularized policy iteration has global linear convergence with the rate being $γ$ (discount factor). Furthermore, this algorithm converges quadratically once it enters a local region around the optimal value. We also show that a modified version of regularized policy iteration, i.e., with finite-step policy evaluation, is equivalent to inexact Newton method where the Newton iteration formula is solved with truncated iterations. We prove that the associated algorithm achieves an asymptotic linear convergence rate of $γ^M$ in which $M$ denotes the number of steps carried out in policy evaluation. Our results take a solid step towards a better understanding of the convergence properties of regularized policy iteration algorithms.

Chenyu Zhu, Zeyang Li, Ziyi Xie et al.

Specific emitter identification (SEI) utilizes passive hardware characteristics to authenticate transmitters, providing a robust physical-layer security solution. However, most deep-learning-based methods rely on extensive data or require prior information, which poses challenges in real-world scenarios with limited labeled data. We propose an integrated complex variational mode decomposition algorithm that decomposes and reconstructs complex-valued signals to approximate the original transmitted signals, thereby enabling more accurate feature extraction. We further utilize a temporal convolutional network to effectively model the sequential signal characteristics, and introduce a spatial attention mechanism to adaptively weight informative signal segments, significantly enhancing identification performance. Additionally, the branch network allows leveraging pre-trained weights from other data while reducing the need for auxiliary datasets. Ablation experiments on the simulated data demonstrate the effectiveness of each component of the model. An accuracy comparison on a public dataset reveals that our method achieves 96% accuracy using only 10 symbols without requiring any prior knowledge.

Yahan Li, Chaohao Du, Zeyang Li et al.

Mental-health support is increasingly mediated by conversational systems (e.g., LLM-based tools), but users often lack structured ways to audit the quality and potential risks of the support they receive. We introduce CounselReflect, an end-to-end toolkit for auditing mental-health support dialogues. Rather than producing a single opaque quality score, CounselReflect provides structured, multi-dimensional reports with session-level summaries, turn-level scores, and evidence-linked excerpts to support transparent inspection. The system integrates two families of evaluation signals: (i) 12 model-based metrics produced by task-specific predictors, and (ii) rubric-based metrics that extend coverage via a literature-derived library (69 metrics) and user-defined custom metrics, operationalized with configurable LLM judges. CounselReflect is available as a web application, browser extension, and command-line interface (CLI), enabling use in real-time settings as well as at scale. Human evaluation includes a user study with 20 participants and an expert review with 6 mental-health professionals, suggesting that CounselReflect supports understandable, usable, and trustworthy auditing. A demo video and full source code are also provided.

Zeyang Li, Navid Azizan · mit

Multi-agent reinforcement learning (MARL) has achieved notable success in cooperative tasks, demonstrating impressive performance and scalability. However, deploying MARL agents in real-world applications presents critical safety challenges. Current safe MARL algorithms are largely based on the constrained Markov decision process (CMDP) framework, which enforces constraints only on discounted cumulative costs and lacks an all-time safety assurance. Moreover, these methods often overlook the feasibility issue (the system will inevitably violate state constraints within certain regions of the constraint set), resulting in either suboptimal performance or increased constraint violations. To address these challenges, we propose a novel theoretical framework for safe MARL with $\textit{state-wise}$ constraints, where safety requirements are enforced at every state the agents visit. To resolve the feasibility issue, we leverage a control-theoretic notion of the feasible region, the controlled invariant set (CIS), characterized by the safety value function. We develop a multi-agent method for identifying CISs, ensuring convergence to a Nash equilibrium on the safety value function. By incorporating CIS identification into the learning process, we introduce a multi-agent dual policy iteration algorithm that guarantees convergence to a generalized Nash equilibrium in state-wise constrained cooperative Markov games, achieving an optimal balance between feasibility and performance. Furthermore, for practical deployment in complex high-dimensional systems, we propose $\textit{Multi-Agent Dual Actor-Critic}$ (MADAC), a safe MARL algorithm that approximates the proposed iteration scheme within the deep RL paradigm. Empirical evaluations on safe MARL benchmarks demonstrate that MADAC consistently outperforms existing methods, delivering much higher rewards while reducing constraint violations.

Shanbo Cheng, Yu Bao, Zhichao Huang et al.

Simultaneous Interpretation (SI) represents one of the most daunting frontiers in the translation industry, with product-level automatic systems long plagued by intractable challenges: subpar transcription and translation quality, lack of real-time speech generation, multi-speaker confusion, and translated speech inflation, especially in long-form discourses. In this study, we introduce Seed-LiveInterpret 2.0, an end-to-end SI model that delivers high-fidelity, ultra-low-latency speech-to-speech generation with voice cloning capabilities. As a fully operational product-level solution, Seed-LiveInterpret 2.0 tackles these challenges head-on through our novel duplex speech-to-speech understanding-generating framework. Experimental results demonstrate that through large-scale pretraining and reinforcement learning, the model achieves a significantly better balance between translation accuracy and latency, validated by human interpreters to exceed 70% correctness in complex scenarios. Notably, Seed-LiveInterpret 2.0 outperforms commercial SI solutions by significant margins in translation quality, while slashing the average latency of cloned speech from nearly 10 seconds to a near-real-time 3 seconds, which is around a near 70% reduction that drastically enhances practical usability.