Institution profile

University of Bristol

Academic institutioneurope · gb
Official website
Research library636linked papers
Opportunities0open roles
Selected work

Representative Papers

A Survey on Vision-Language-Action Models for Embodied AI

May 23, 2024arXiv.org

This paper addresses the core challenge of how Vision-Language-Action (VLA) models support language-conditioned robotic tasks in embodied intelligence. Methodologically, it introduces the first systematic, panoramic survey framework, proposing a three-dimensional taxonomy—“Component Design–Low-level Action Policies–High-level Task Planning”—that unifies VLA modeling, embodied control, task decomposition, simulation integration, and cross-benchmark evaluation. Key contributions include: (1) the first explicit characterization of three principal VLA technical paradigms; (2) a comprehensive survey of multimodal datasets, embodied simulation platforms, and standardized evaluation benchmarks; and (3) a structured knowledge graph that identifies critical open challenges—including scalable architecture design, world model integration, and real-world deployment—and outlines promising future research directions.

18 citations1 influentialRead paper

Game Connectivity and Adaptive Dynamics

Sep 19, 2023arXiv.org

This paper investigates the connectivity of best-response graphs in generic games: Why, in large random games admitting pure Nash equilibria (PNE), do almost all non-equilibrium strategy profiles admit a best-response path to some PNE? Method: Leveraging an interdisciplinary approach combining probabilistic combinatorics, game theory, and graph theory, the authors establish the first universal probabilistic relationship among game size, existence of PNE, and best-response graph connectivity. Contribution/Results: They prove that for almost all large generic games possessing at least one PNE, the associated best-response graph is asymptotically almost surely connected. This graph-theoretic connectivity serves as a sufficient structural condition ensuring almost-sure convergence of adaptive dynamics—such as inertia-augmented best-response processes—to a PNE. The result uncovers an intrinsic robustness in equilibrium accessibility within high-dimensional games and provides a foundational topological explanation for the convergence of uncoupled learning mechanisms.

5 citationsRead paper

Policy choice in time series by empirical welfare maximization

May 08, 2022

Dynamic multivariate time series pose challenges including time-varying environments, historical dependence, dynamic causal effects, and statistical dependencies. Method: This paper proposes Time-series Empirical Welfare Maximization (T-EWM), the first extension of the empirical welfare maximization framework to dynamic time-series settings. T-EWM employs nonparametric potential outcome modeling and conditional welfare optimization to learn dynamic optimal policies. Contribution/Results: We establish theoretical guarantees, including conditional welfare consistency and a non-asymptotic upper bound on policy regret. In simulation studies and an empirical application to COVID-19 containment policy evaluation, T-EWM significantly improves policy welfare and achieves rapid regret convergence under limited samples. The framework provides a novel paradigm for dynamic decision-making that balances statistical rigor with practical feasibility.

4 citationsRead paper

Multi-Objective Pareto-Front Optimization for Efficient Adaptive VVC Streaming

Jan 15, 2026

This work addresses the challenge of adaptive video streaming by balancing bitrate, visual quality, and decoding complexity while accounting for content characteristics and ensuring consistent user experience—limitations not adequately met by existing approaches. The authors propose a multi-objective Pareto-front optimization framework that, for the first time, incorporates decoding time as a proxy for energy consumption. Within this framework, two strategies—JRQT-PF and JQT-PF—are introduced to generate content-adaptive, efficient bitrate ladders for VVC under a quality monotonicity constraint. Experimental results demonstrate that JQT-PF achieves an average bitrate saving of 11.76% with slightly reduced decoding time while maintaining XPSNR, whereas JRQT-PF yields 6.38% bitrate savings and a 6.17% reduction in decoding time, significantly outperforming both fixed-ladder and dynamic-resolution baselines.

3 citationsRead paper

Systemic Risk in DeFi: A Network-Based Fragility Analysis of TVL Dynamics

Jan 13, 2026

This study addresses the lack of structured, continuous systemic risk assessment in decentralized finance (DeFi), stemming from its highly interconnected nature. To this end, the authors propose a unified quantitative framework that integrates time-varying correlation networks with the dynamics of Total Value Locked (TVL). By constructing a time-evolving network of DeFi protocols and combining complex network analysis with modular functional classification, they introduce two novel metrics: the Correlation-based Fragility Index (CFI) and the Risk Contribution Score (RCS). This approach overcomes the limitations of traditional event-driven or single-channel analyses, enabling dynamic tracking of systemic risk and effectively identifying protocol categories that play pivotal roles in risk accumulation and amplification.

1 citationsRead paper
Recent publications

Latest Papers