Selected work

Representative Papers

Matching random colored points with rectangles (Corrigendum)

Mar 31, 2025

This paper studies the problem of matching $n$ points sampled uniformly at random in the unit square and colored red or blue, using pairwise disjoint axis-aligned rectangles to match points of the same color, with the goal of maximizing the number $M(n)$ of covered points. Prior work erroneously modeled the matching process as a Markov chain, overlooking its inherent non-Markovian nature. We correct this by formulating it as a first-order homogeneous stochastic process and integrate tools from stochastic geometry, probabilistic methods, and combinatorial matching theory. Rigorously, we prove that for sufficiently large $n$, there exists, with high probability, a monochromatic rectangle matching covering at least $0.83n$ points—i.e., $M(n) geq 0.83n$. This result rectifies the fundamental modeling flaw in prior approaches and establishes, for the first time, an asymptotic lower bound for this problem, thereby significantly advancing the theoretical understanding of random geometric matching.

1 citationsRead paper

Risk Assessment of Transmission Lines Against Grid-ignited Wildfires

Feb 18, 2025

Under increasing wildfire frequency, systematic assessment of transmission line ignition risk—and its cascading environmental and grid impacts—is urgently needed. This paper proposes the first multidimensional risk assessment framework integrating line ignition potential, regional environmental vulnerability, and power grid topological propagation effects, enabled by multisource fusion of meteorological, geographical, and grid topology data. Its key innovation lies in unifying the quantification of physical ignition mechanisms, ecological/urban exposure, and fault propagation pathways to support differentiated disaster mitigation strategy prioritization. Validated on the IEEE 30-bus system, the framework accurately identifies high-risk transmission lines and provides quantifiable risk metrics to inform vegetation management, preemptive de-energization decisions, and undergrounding investments. Results demonstrate significant enhancement in power grid wildfire resilience through data-driven, risk-informed infrastructure planning and operational response.

1 citationsRead paper

SAT-Based Techniques for Lexicographically Smallest Finite Models

Mar 24, 2024AAAI Conference on Artificial Intelligence

This paper addresses the isomorphism identification and canonical labeling problem for finite algebraic structures (e.g., groups, semigroups). We propose a SAT-based lexicographic normalization method that computes the lexicographically smallest representation of a structure via domain element reordering. Our approach employs a black-box SAT framework to progressively construct the minimal representative and introduces a novel constraint propagation mechanism that substantially reduces the number of SAT solver invocations. Unlike prior approaches, it supports arbitrary finite algebraic structures and achieves fully automated lexicographic canonicalization for the first time. We implement an open-source tool that processes structures of order up to 1,000 in real time. Empirical evaluation demonstrates significant speedups over naive enumeration—reducing both total solving time and SAT calls—while maintaining correctness and scalability across diverse algebraic classes, including groups, semigroups, and magmas.

1 citationsRead paper

Language-Based Digital Twins for Elderly Cognitive Assistance

Jun 25, 2026

This study addresses the challenge of early detection of mild cognitive impairment (MCI) in older adults, hindered by the lack of non-invasive, personalized, and continuous monitoring tools. The authors propose the first language-based digital twin framework that leverages large language models to simulate individual conversational behavior by integrating stylistic features with contextual metadata, thereby constructing a personalized cognitive health model. The core innovation lies in a novel multi-head conditional variational autoencoder designed to jointly optimize identity fidelity and cognitive consistency. Experiments on the I-CONECT dataset demonstrate that the proposed method matches real conversations in terms of speaker identity preservation, dialogue reconstruction quality, and MoCA score prediction accuracy, significantly outperforming baseline generative approaches such as GPT.

0 citationsRead paper

Comment on "The Forsaken Road: Reassessing Living Standards Following the Cuban Revolution and the American Embargo"

Apr 21, 2026

This study reevaluates the actual impact of the U.S. embargo on post-revolutionary Cuba’s per capita income divergence, challenging prior findings that downplay the embargo’s role. By revising the parameterization of trade openness elasticity with respect to income and adopting elasticity values and interaction effect decomposition methods more consistent with the scholarly consensus, the analysis integrates counterfactual reasoning with econometric techniques to reassess the relative contributions of the embargo and other growth determinants. The results indicate that the U.S. embargo accounts for a substantial share of Cuba’s underperformance since 1959 and, under certain scenarios, is sufficient to fully explain its persistent economic lag, thereby significantly enhancing the explanatory power attributed to the embargo.

0 citationsRead paper

Recent publications

Latest Papers

Language-Based Digital Twins for Elderly Cognitive Assistance

Jun 25, 2026

This study addresses the challenge of early detection of mild cognitive impairment (MCI) in older adults, hindered by the lack of non-invasive, personalized, and continuous monitoring tools. The authors propose the first language-based digital twin framework that leverages large language models to simulate individual conversational behavior by integrating stylistic features with contextual metadata, thereby constructing a personalized cognitive health model. The core innovation lies in a novel multi-head conditional variational autoencoder designed to jointly optimize identity fidelity and cognitive consistency. Experiments on the I-CONECT dataset demonstrate that the proposed method matches real conversations in terms of speaker identity preservation, dialogue reconstruction quality, and MoCA score prediction accuracy, significantly outperforming baseline generative approaches such as GPT.

0 citationsRead paper

Comment on "The Forsaken Road: Reassessing Living Standards Following the Cuban Revolution and the American Embargo"

Apr 21, 2026

This study reevaluates the actual impact of the U.S. embargo on post-revolutionary Cuba’s per capita income divergence, challenging prior findings that downplay the embargo’s role. By revising the parameterization of trade openness elasticity with respect to income and adopting elasticity values and interaction effect decomposition methods more consistent with the scholarly consensus, the analysis integrates counterfactual reasoning with econometric techniques to reassess the relative contributions of the embargo and other growth determinants. The results indicate that the U.S. embargo accounts for a substantial share of Cuba’s underperformance since 1959 and, under certain scenarios, is sufficient to fully explain its persistent economic lag, thereby significantly enhancing the explanatory power attributed to the embargo.

0 citationsRead paper

Contrastive learning-based video quality assessment-jointed video vision transformer for video recognition

Mar 11, 2026

This work addresses the challenge that video blur significantly degrades classification performance, while no-reference video quality assessment (VQA) remains difficult to model due to the absence of ground-truth labels. To this end, the authors propose SSL-V3, a novel approach that integrates no-reference VQA into the video classification pipeline via a joint self-supervised learning framework. By leveraging the classification task to inversely optimize VQA parameters and dynamically modulating classification features with estimated quality scores, SSL-V3 enables quality-aware robust recognition without requiring VQA ground-truth annotations. Built upon a contrastive learning-based Video Vision Transformer, the method demonstrates strong empirical performance, achieving a classification accuracy of 94.87% on interview videos in the I-CONECT dataset and validating its effectiveness across multiple benchmarks.

0 citationsRead paper

Predicting Atomistic Transitions with Transformers

Mar 05, 2026

This work addresses the high computational cost of conventional atomistic transition path sampling, which has hindered its practical application. For the first time, the Transformer architecture is introduced into this domain to construct an efficient surrogate model capable of learning complex emergent behaviors from atomic simulation data, enabling rapid prediction of atomic transition pathways in nanoclusters. The proposed method supports fine-tuning of inputs to generate diverse yet physically plausible microstates, with built-in validation of physical consistency to ensure reliability. Experimental results demonstrate that the model substantially reduces computational overhead while maintaining strong generalization performance and adherence to physical principles.

0 citationsRead paper

CodeEval: A pedagogical approach for targeted evaluation of code-trained Large Language Models

Jan 06, 2026arXiv.org

This work addresses the limitation of existing benchmarks for code-generating large language models, which lack fine-grained assessment of programming proficiency and thus hinder targeted model improvement. To this end, we propose a pedagogy-inspired, multidimensional evaluation framework and introduce CodeEval—a benchmark encompassing 24 Python programming dimensions across three difficulty levels—accompanied by the open-source RunCodeEval execution framework. Our approach uniquely integrates assessment principles from academic programming curricula into large model evaluation, enabling granular diagnostic capabilities through function- and class-level tasks, tiered test cases, and automated metric generation. This system supports fine-grained analysis of model performance across difficulty levels, problem types, and programming constructs. Designed for reproducibility and ease of deployment, the framework empowers researchers to efficiently identify model strengths and weaknesses in specific programming skills, thereby facilitating directed enhancements.

0 citationsRead paper