arXiv subjects

Xiang Yin

Publications and source records attributed to Xiang Yin.

At least 55 records · Page 3Linked to original sources

arXiv2026

ALIVE: Animate Your World with Lifelike Audio-Video Generation

Video generation is rapidly evolving towards unified audio-video generation. In this paper, we present ALIVE, a generation model that adapts a pretrained Text-to-Video (T2V) model to Sora-style audio-video generation and animation. In particular, the model unlocks the Text-to-Video&Audio (T2VA) and Reference-to-Video&Audio (animation) capabilities compared to the T2V foundation models. To support the audio-visual synchronization and reference animation, we augment the popular MMDiT architecture with a joint audio-video branch which includes TA-CrossAttn for temporally-aligned cross-modal fusion and UniTemp-RoPE for precise audio-visual alignment. Meanwhile, a comprehensive data pipeline consisting of audio-video captioning, quality control, etc., is carefully designed to collect high-quality finetuning data. Additionally, we introduce a new benchmark to perform a comprehensive model test and comparison. After continue pretraining and finetuning on million-level high-quality data, ALIVE demonstrates outstanding performance, consistently outperforming open-source models and matching or surpassing state-of-the-art commercial solutions. With detailed recipes and benchmarks, we hope ALIVE helps the community develop audio-video generation models more efficiently. Official page: https://github.com/FoundationVision/Alive.

cs.CV↗

arXiv2026

Latent Debate: A Surrogate Framework for Interpreting LLM Thinking

Understanding the internal thinking process of Large Language Models (LLMs) and the cause of hallucinations remains a key challenge. To this end, we introduce latent debate, a novel framework for interpreting model predictions through the lens of implicit internal arguments. Unlike the current work of self-consistency and multi-agent debate, which relies on explicit debates among multiple answers or multiple models, latent debate captures the hidden supporting and attacking signals that arise within a single model during a single inference. We first present a model- and task-agnostic conceptual framework, and then instantiate it symbolically to approximate the thinking process of LLMs on True/False prediction tasks. Empirical studies demonstrate that latent debate is a faithful structured surrogate model that has highly consistent predictions with the original LLM. Beyond interpretability, we demonstrate that latent debate provides a strong baseline for hallucination detection. Further analysis reveals strong correlations between hallucinations and debate patterns, such as a high degree of latent debates in the middle layers is linked to a higher risk of hallucinations. These findings position latent debate as a potential framework for understanding internal mechanisms of LLMs, especially for scenarios where internal (dis)agreements appear during the inference steps.

cs.CL↗

arXiv2026

Strength Change Explanations in Quantitative Argumentation

In order to make argumentation-based inference contestable, it is crucial to explain what changes can achieve a desired (instead of the contested) inference result. To this end, we introduce strength change explanations for quantitative (bipolar) argumentation graphs. Strength change explanations describe changes to the initial strengths of a subset of the arguments in a given graph that can achieve a desired ordering based on the final strengths of some (potentially different) subset of arguments. We show that the existing notions of inverse and counterfactual problems can be reduced to strength change explanations. We also prove basic soundness and completeness properties of our strength change explanations, and demonstrate their existence and non-existence in some special cases. By applying a heuristic search, we demonstrate that we can often successfully find strength change explanations for layered graphs that are common in typical application scenarios; still, limitations remain for settings where we do not provide guarantees for the presence (or absence) of explanations.

cs.MA↗

arXiv2026

On the Computation and Approximation of Backward Reachable Sets for Max-Plus Linear Systems using Polyhedras

This paper investigates reachability analysis for max-plus linear systems (MPLS), an important class of dynamical systems that model synchronization and delay phenomena in timed discrete-event systems. We specifically focus on backward reachability analysis, i.e., determining the set of states that can reach a given target set within a certain number of steps. Computing backward reachable sets presents significant challenges due to the non-convexity of max-plus dynamics and the complexity of set complement operations. To address these challenges, we propose a novel approximation framework that efficiently computes backward reachable sets by exploiting the structure of tropical polyhedra. Our approach reformulates the problem as a sequence of symbolic operations and approximates non-convex target sets through closure operations on unions of tropical polyhedra. We develop a systematic algorithm that constructs both outer (M-form) and inner (V-form) representations of the resulting sets, incorporating extremal filtering to reduce computational complexity. The proposed method offers a scalable alternative to traditional DBM-based approaches, enabling reliable approximate backward reachability analysis for general target regions in MPLS.

eess.SY↗

arXiv2026

Stability Verification for Switched Systems using Neural Multiple Lyapunov Functions

Stability analysis of switched systems, characterized by multiple operational modes and switching signals, is challenging due to their nonlinear dynamics. While frameworks such as multiple Lyapunov functions (MLF) provide a foundation for analysis, their computational applicability is limited for systems without favorable structure. This paper investigates stability analysis for switched systems under state-dependent switching conditions. We propose neural multiple Lyapunov functions (NMLF), a unified framework that combines the theoretical guarantees of MLF with the computational efficiency of neural Lyapunov functions (NLF). Our approach leverages a set of tailored loss functions and a counter-example guided inductive synthesis (CEGIS) scheme to train neural networks that rigorously satisfy MLF conditions. Through comprehensive simulations and theoretical analysis, we demonstrate NMLF's effectiveness and its potential for practical deployment in complex switched systems.

eess.SY↗

arXiv2025

RRT*former: Environment-Aware Sampling-Based Motion Planning using Transformer

We investigate the sampling-based optimal path planning problem for robotics in complex and dynamic environments. Most existing sampling-based algorithms neglect environmental information or the information from previous samples. Yet, these pieces of information are highly informative, as leveraging them can provide better heuristics when sampling the next state. In this paper, we propose a novel sampling-based planning algorithm, called \emph{RRT*former}, which integrates the standard RRT* algorithm with a Transformer network in a novel way. Specifically, the Transformer is used to extract features from the environment and leverage information from previous samples to better guide the sampling process. Our extensive experiments demonstrate that, compared to existing sampling-based approaches such as RRT*, Neural RRT*, and their variants, our algorithm achieves considerable improvements in both the optimality of the path and sampling efficiency. The code for our implementation is available on https://github.com/fengmingyang666/RRTformer.

cs.RO↗

arXiv2025

Distributionally Robust Control Synthesis for Stochastic Systems with Safety and Reach-Avoid Specifications

We investigate the problem of synthesizing distributionally robust control policies for stochastic systems under safety and reach-avoid specifications. Using a game-theoretical framework, we consider the setting where the probability distribution of the disturbance at each time step is selected from an ambiguity set defined by the Wasserstein distance. The goal is to synthesize a distributionally robust control policy that ensures the satisfaction probability exceeds a specified threshold under any distribution within the ambiguity set. First, for both safety and reach-avoid specifications, we establish the existence of optimal policies by leveraging the dynamic programming principles. Then we demonstrate how the associated optimization problem can be efficiently solved using the dual representation of Wasserstein distributionally robust optimization. Furthermore, for safety specifications in particular, we introduce a novel concept of distributionally robust control barrier certificates and show how these enable the efficient synthesis of controllers through sum-of-squares programming techniques. Finally, our experimental results reveal that incorporating distributional robustness during the synthesis phase significantly improves the satisfaction probability during online execution, even with limited statistical knowledge of the disturbance distribution.

eess.SY↗

arXiv2025

Zero-Shot Trajectory Planning for Signal Temporal Logic Tasks

Signal Temporal Logic (STL) is a powerful specification language for describing complex temporal behaviors of continuous signals, making it well-suited for high-level robotic task descriptions. However, generating executable plans for STL tasks is challenging, as it requires consideration of the coupling between the task specification and the system dynamics. Existing approaches either follow a model-based setting that explicitly requires knowledge of the system dynamics or adopt a task-oriented data-driven approach to learn plans for specific tasks. In this work, we address the problem of generating executable STL plans for systems with unknown dynamics. We propose a hierarchical planning framework that enables zero-shot generalization to new STL tasks by leveraging only task-agnostic trajectory data during offline training. The framework consists of three key components: (i) decomposing the STL specification into several progresses and time constraints, (ii) searching for timed waypoints that satisfy all progresses under time constraints, and (iii) generating trajectory segments using a pre-trained diffusion model and stitching them into complete trajectories. We formally prove that our method guarantees STL satisfaction, and simulation results demonstrate its effectiveness in generating dynamically feasible trajectories across diverse long-horizon STL tasks.

cs.RO↗

arXiv2025

On Prediction-Based Properties of Discrete-Event Systems: Notions, Applications and Supervisor Synthesis

In this work, we investigate the problem of synthesizing property-enforcing supervisors for partially-observed discrete-event systems (DES). Unlike most existing approaches, where the enforced property depends solely on the executed behavior of the system, here we consider a more challenging scenario in which the property relies on predicted future behaviors that have not yet occurred. This problem arises naturally in applications involving future information, such as active prediction or intention protection. To formalize the problem, we introduce the notion of prediction-based properties, a new class of observational properties tied to the system's future information. We demonstrate that this notion is very generic and can model various practical properties, including predictability in fault prognosis and pre-opacity in intention security. We then present an effective approach for synthesizing supervisors that enforce prediction-based properties. Our method relies on a novel information structure that addresses the fundamental challenge arising from the dependency between current predictions and the control policy. The key idea is to first borrow information from future instants and then ensure information consistency. This reduces the supervisor synthesis problem to a safety game in the information space. We prove that the proposed algorithm is both sound and complete, and the resulting supervisor is maximally permissive.

eess.SY↗

arXiv2025

Optimal Control of Markov Decision Processes for Efficiency with Linear Temporal Logic Tasks

We investigate the problem of optimal control synthesis for Markov Decision Processes (MDPs), addressing both qualitative and quantitative objectives. Specifically, we require the system to satisfy a qualitative task specified by a Linear Temporal Logic (LTL) formula with probability one. Additionally, to quantify the system's performance, we introduce the concept of efficiency, defined as the ratio between rewards and costs. This measure is more general than the standard long-run average reward metric, as it seeks to maximize the reward obtained per unit cost. Our objective is to synthesize a control policy that not only ensures the LTL task is satisfied but also maximizes efficiency. We present an effective approach for synthesizing a stationary control policy that achieves $ε$-optimality by integrating state classifications of MDPs with perturbation analysis in a novel manner. Our results extend existing work on efficiency-optimal control synthesis for MDPs by incorporating qualitative LTL tasks. Case studies in robot task planning are provided to illustrate the proposed algorithm.

eess.SY↗

arXiv2025

VH-Diffuser: Variable Horizon Diffusion Planner for Time-Aware Goal-Conditioned Trajectory Planning

Diffusion-based planners have gained significant recent attention for their robustness and performance in long-horizon tasks. However, most existing planners rely on a fixed, pre-specified horizon during both training and inference. This rigidity often produces length-mismatch (trajectories that are too short or too long) and brittle performance across instances with varying geometric or dynamical difficulty. In this paper, we introduce the Variable Horizon Diffuser (VHD) framework, which treats the horizon as a learned variable rather than a fixed hyperparameter. Given a start-goal pair, we first predict an instance-specific horizon using a learned Length Predictor model, which guides a Diffusion Planner to generate a trajectory of the desired length. Our design maintains compatibility with existing diffusion planners by controlling trajectory length through initial noise shaping and training on randomly cropped sub-trajectories, without requiring architectural changes. Empirically, VHD improves success rates and path efficiency in maze-navigation and robot-arm control benchmarks, showing greater robustness to horizon mismatch and unseen lengths, while keeping training simple and offline-only.

cs.RO↗

arXiv2025

Control Synthesis for Multiple Reach-Avoid Tasks via Hamilton-Jacobi Reachability Analysis

We investigate the control synthesis problem for continuous-time time-varying nonlinear systems with disturbance under a class of multiple reach-avoid (MRA) tasks. Specifically, the MRA task requires the system to reach a series of target regions in a specified order while satisfying state constraints between each pair of target arrivals. This problem is more challenging than standard reach-avoid tasks, as it requires considering the feasibility of future reach-avoid tasks during the planning process. To solve this problem, we define a series of value functions by solving a cascade of time-varying reach-avoid problems characterized by Hamilton-Jacobi variational inequalities. We prove that the super-level set of the final value function computed is exactly the feasible set of the MRA task. Additionally, we demonstrate that the control law can be effectively synthesized by ensuring the non-negativeness of the value functions over time. We also show that the Linear temporal logic task control synthesis problems can be converted to a collection of MRA task control synthesis problems by properly defining each target and state constraint set of MRA tasks. The effectiveness of the proposed approach is illustrated through four case studies on robot planning problems under time-varying nonlinear systems with disturbance.

eess.SY↗

arXiv2025

Task and Motion Planning of Dynamic Systems using Hyperproperties for Signal Temporal Logics

We investigate the task and motion planning problem for dynamical systems under signal temporal logic (STL) specifications. Existing works on STL control synthesis mainly focus on generating plans that satisfy properties over a single executed trajectory. In this work, we consider the planning problem for hyperproperties evaluated over a set of possible trajectories, which naturally arise in information-flow control problems. Specifically, we study discrete-time dynamical systems and employ the recently developed temporal logic HyperSTL as the new objective for planning. To solve this problem, we propose a novel recursive counterexample-guided synthesis approach capable of effectively handling HyperSTL specifications with multiple alternating quantifiers. The proposed method is not only applicable to planning but also extends to HyperSTL model checking for discrete-time dynamical systems. Finally, we present case studies on security-preserving planning and ambiguity-free planning to demonstrate the effectiveness of the proposed HyperSTL planning framework.

eess.SY↗

arXiv2025

InfinityHuman: Towards Long-Term Audio-Driven Human

Audio-driven human animation has attracted wide attention thanks to its practical applications. However, critical challenges remain in generating high-resolution, long-duration videos with consistent appearance and natural hand motions. Existing methods extend videos using overlapping motion frames but suffer from error accumulation, leading to identity drift, color shifts, and scene instability. Additionally, hand movements are poorly modeled, resulting in noticeable distortions and misalignment with the audio. In this work, we propose InfinityHuman, a coarse-to-fine framework that first generates audio-synchronized representations, then progressively refines them into high-resolution, long-duration videos using a pose-guided refiner. Since pose sequences are decoupled from appearance and resist temporal degradation, our pose-guided refiner employs stable poses and the initial frame as a visual anchor to reduce drift and improve lip synchronization. Moreover, to enhance semantic accuracy and gesture realism, we introduce a hand-specific reward mechanism trained with high-quality hand motion data. Experiments on the EMTD and HDTF datasets show that InfinityHuman achieves state-of-the-art performance in video quality, identity preservation, hand accuracy, and lip-sync. Ablation studies further confirm the effectiveness of each module. Code will be made public.

cs.CV↗

arXiv2025

UniTalker: Conversational Speech-Visual Synthesis

Conversational Speech Synthesis (CSS) is a key task in the user-agent interaction area, aiming to generate more expressive and empathetic speech for users. However, it is well-known that "listening" and "eye contact" play crucial roles in conveying emotions during real-world interpersonal communication. Existing CSS research is limited to perceiving only text and speech within the dialogue context, which restricts its effectiveness. Moreover, speech-only responses further constrain the interactive experience. To address these limitations, we introduce a Conversational Speech-Visual Synthesis (CSVS) task as an extension of traditional CSS. By leveraging multimodal dialogue context, it provides users with coherent audiovisual responses. To this end, we develop a CSVS system named UniTalker, which is a unified model that seamlessly integrates multimodal perception and multimodal rendering capabilities. Specifically, it leverages a large-scale language model to comprehensively understand multimodal cues in the dialogue context, including speaker, text, speech, and the talking-face animations. After that, it employs multi-task sequence prediction to first infer the target utterance's emotion and then generate empathetic speech and natural talking-face animations. To ensure that the generated speech-visual content remains consistent in terms of emotion, content, and duration, we introduce three key optimizations: 1) Designing a specialized neural landmark codec to tokenize and reconstruct facial expression sequences. 2) Proposing a bimodal speech-visual hard alignment decoding strategy. 3) Applying emotion-guided rendering during the generation stage. Comprehensive objective and subjective experiments demonstrate that our model synthesizes more empathetic speech and provides users with more natural and emotionally consistent talking-face animations.

eess.AS↗

arXiv2025

HumanDiT: Pose-Guided Diffusion Transformer for Long-form Human Motion Video Generation

Human motion video generation has advanced significantly, while existing methods still struggle with accurately rendering detailed body parts like hands and faces, especially in long sequences and intricate motions. Current approaches also rely on fixed resolution and struggle to maintain visual consistency. To address these limitations, we propose HumanDiT, a pose-guided Diffusion Transformer (DiT)-based framework trained on a large and wild dataset containing 14,000 hours of high-quality video to produce high-fidelity videos with fine-grained body rendering. Specifically, (i) HumanDiT, built on DiT, supports numerous video resolutions and variable sequence lengths, facilitating learning for long-sequence video generation; (ii) we introduce a prefix-latent reference strategy to maintain personalized characteristics across extended sequences. Furthermore, during inference, HumanDiT leverages Keypoint-DiT to generate subsequent pose sequences, facilitating video continuation from static images or existing videos. It also utilizes a Pose Adapter to enable pose transfer with given sequences. Extensive experiments demonstrate its superior performance in generating long-form, pose-accurate videos across diverse scenarios.

cs.CV↗

arXiv2025

On the Construction of Barrier Certificate: A Dynamic Programming Perspective

In this paper, we revisit the formal verification problem for stochastic dynamical systems over finite horizon using barrier certificates. Most existing work on this topic focuses on safety properties by constructing barrier certificates based on the notion of $c$-martingales. In this work, we first provide a new insight into the conditions of existing martingale-based barrier certificates from the perspective of dynamic programming operators. Specifically, we show that the existing conditions essentially provide a bound on the dynamic programming solution, which exactly characterizes the safety probability. Based on this new perspective, we demonstrate that the barrier conditions in existing approaches are unnecessarily conservative over unsafe states. To address this, we propose a new set of safety barrier certificate conditions that are strictly less conservative than existing ones, thereby providing tighter probability bounds for safety verification. We further extend our approach to the case of reach-avoid specifications by providing a set of new barrier certificate conditions. We also illustrate how to search for these new barrier certificates using sum-of-squares (SOS) programming. Finally, we use two numerical examples to demonstrate the advantages of our method compared to existing approaches.

eess.SY↗

arXiv2025

Signal Temporal Logic Control Synthesis among Uncontrollable Dynamic Agents with Conformal Prediction

The control of dynamical systems under temporal logic specifications among uncontrollable dynamic agents is challenging due to the agents' a-priori unknown behavior. Existing works have considered the problem where either all agents are controllable, the agent models are deterministic and known, or no safety guarantees are provided. We propose a predictive control synthesis framework that guarantees, with high probability, the satisfaction of signal temporal logic (STL) tasks that are defined over a controllable system in the presence of uncontrollable stochastic agents. We use trajectory predictors and conformal prediction to construct probabilistic prediction regions for each uncontrollable agent that are valid over multiple future time steps. Specifically, we construct a normalized prediction region over all agents and time steps to reduce conservatism and increase data efficiency. We then formulate a worst-case bilevel mixed integer program (MIP) that accounts for all agent realizations within the prediction region to obtain an open-loop controller that provably guarantee task satisfaction with high probability. To efficiently solve this bilevel MIP, we propose an equivalent MIP program based on KKT conditions of the original bilevel formulation. Building upon this, we design a closed-loop controller, where both recursive feasibility and task satisfaction can be guaranteed with high probability. We illustrate our control synthesis framework on two case studies.

eess.SY↗