Designing Agentic AI Workflow Portfolios under Imperfect Selection and Compute Cost
研究通过组合多种AI工作流并选择最佳结果的方法，以解决单一最优工作流可能错过正确答案的问题，使用线性规划等技术优化组合。
0 citationsRead paper
Selected work
Representative Papers
Illustrating Hyperbolic Surfaces with Mesh Embeddings
为解决双曲几何在欧氏空间中难以直观展示的问题，本文通过将双曲表面离散化为网格，并优化变形能量以匹配双曲平面中的边长，从而在欧氏空间中嵌入双曲表面。
0 citationsRead paper
Multilingual Agent System for Inclusive Wildfire Evacuation Guidance
为解决野火疏散信息不均问题，开发了BEACON系统，通过多语言聊天机器人、个性化清单等提供实时避险指导。
0 citationsRead paper
Scientific Visualization as a Collaborative Data Infrastructure
本文通过回顾NASA团队使用PIXLISE平台进行火星地质研究的案例，探讨了科学可视化作为协作数据基础设施在数据处理和证据呈现中的作用及未来潜力。
0 citationsRead paper
Connectivity of Districting Metagraphs
This study investigates the irreducibility of ReCom-based Markov chains in political redistricting to ensure representative sampling of districting plans. Employing tools from graph theory and Markov chain Monte Carlo methods, the authors establish—for the first time—the irreducibility of the two-partition ReCom chain on subgraphs of the triangular lattice and prove irreducibility for three-equal-partition ReCom chains on 3×n grid graphs. Conversely, they construct an infinite family of counterexamples demonstrating that even minor structural variations in closely related planar subdivisions can render the chain disconnected, highlighting the fragility of irreducibility. These results delineate critical theoretical boundaries for assessing fairness in redistricting algorithms.
0 citationsRead paper
Recent publications
Latest Papers
Designing Agentic AI Workflow Portfolios under Imperfect Selection and Compute Cost
研究通过组合多种AI工作流并选择最佳结果的方法，以解决单一最优工作流可能错过正确答案的问题，使用线性规划等技术优化组合。
0 citationsRead paper
Illustrating Hyperbolic Surfaces with Mesh Embeddings
为解决双曲几何在欧氏空间中难以直观展示的问题，本文通过将双曲表面离散化为网格，并优化变形能量以匹配双曲平面中的边长，从而在欧氏空间中嵌入双曲表面。
0 citationsRead paper
Multilingual Agent System for Inclusive Wildfire Evacuation Guidance
为解决野火疏散信息不均问题，开发了BEACON系统，通过多语言聊天机器人、个性化清单等提供实时避险指导。
0 citationsRead paper
Scientific Visualization as a Collaborative Data Infrastructure
本文通过回顾NASA团队使用PIXLISE平台进行火星地质研究的案例，探讨了科学可视化作为协作数据基础设施在数据处理和证据呈现中的作用及未来潜力。
0 citationsRead paper
Connectivity of Districting Metagraphs
This study investigates the irreducibility of ReCom-based Markov chains in political redistricting to ensure representative sampling of districting plans. Employing tools from graph theory and Markov chain Monte Carlo methods, the authors establish—for the first time—the irreducibility of the two-partition ReCom chain on subgraphs of the triangular lattice and prove irreducibility for three-equal-partition ReCom chains on 3×n grid graphs. Conversely, they construct an infinite family of counterexamples demonstrating that even minor structural variations in closely related planar subdivisions can render the chain disconnected, highlighting the fragility of irreducibility. These results delineate critical theoretical boundaries for assessing fairness in redistricting algorithms.
0 citationsRead paper