Selected work

Representative Papers

TriView-YOLO: Early Multi-View Fusion for Ground Penetrating Radar Cavity Detection in Soft, High-Water-Content Soils

Aug 10, 2026

This study addresses the challenge of severe signal attenuation and poor void detectability in ground-penetrating radar (GPR) data acquired in high-moisture soft soils by proposing TriView-YOLO, a novel detection model based on the YOLOv12 architecture. The method introduces, for the first time, a nine-channel input comprising three orthogonal views—vertical B-scan, horizontal C-scan, and cross-sectional B-scan—and incorporates a custom TripleInputConv layer to enable early multi-view fusion, while producing detection bounding boxes exclusively from the vertical view. Trained on real-world data collected via a vehicle-mounted 3D GPR system, the model is evaluated under a rigorously designed protocol tailored to complex field conditions lacking public benchmarks. Experimental results demonstrate a mean average precision (mAP50) of 0.558 ± 0.028 on non-augmented test data, with per-image inference requiring only 3.1 ms (23.6 GFLOPs), significantly outperforming ablated variants and confirming the critical contribution of auxiliary views to detection performance.

0 citationsRead paper

Privacy-Preserving Product-Quantized Approximate Nearest Neighbor Search Framework for Large-scale Datasets via A Hybrid of Fully Homomorphic Encryption and Trusted Execution Environment

Apr 20, 2026

Existing privacy-preserving approximate nearest neighbor (PP-ANN) methods struggle to balance security and efficiency and often overlook privacy risks during database construction, leaving them vulnerable to embedding inversion and membership inference attacks. This work proposes PPPQ-ANN, the first framework to synergistically integrate fully homomorphic encryption (FHE) and trusted execution environments (TEEs) across the entire ANN pipeline—spanning both database construction and query processing. By combining product quantization with ciphertext packing optimizations, the framework substantially reduces FHE computational overhead. Evaluated on million-scale datasets, the system achieves database construction in under two hours and sequential search throughput exceeding 50 queries per second, demonstrating practical performance while providing strong privacy guarantees.

0 citationsRead paper

From Minimal Existence to Human Definition: The CES-IMU-HSG Theoretical Framework

Oct 15, 2025

This paper addresses the lack of ontological autonomy in artificial intelligence by proposing a cross-universe mathematical logical framework grounded in Descartes’ “cogito, ergo sum” (CES) as a minimal existence axiom. Methodologically, it integrates an intermediate metaverse (IMU) with a hierarchical state grid (HSG), introduces the categorical identity “definition = state”, and couples neuro-endocrine-genetic physiological systems via category theory, Institution theory, and fibration structures to enable multi-scale modeling over 0–3D neural functional fields. The core contribution is twofold: first, it formalizes human cognition—constrained by physical embodiment—as a cross-universe temporal algorithm; second, it endows machines with an intrinsic CES axiom, thereby establishing a formally rigorous, self-referential foundation for existence. This bridges philosophical ontology and AI engineering autonomy through a mathematically continuous, ontologically grounded framework.

0 citationsRead paper

Combined Hyperbolic and Euclidean Soft Triple Loss Beyond the Single Space Deep Metric Learning

Oct 07, 2025

Existing deep metric learning (DML) methods lack proxy-based supervised losses in hyperbolic space, limiting scalability to large-scale data and semantic modeling capability. To address this, we propose the first dual-space joint optimization framework that simultaneously learns in both hyperbolic and Euclidean spaces. Our approach introduces the first hyperbolic proxy loss, coupled with a novel dual-space soft triplet loss and a hyperbolic hierarchical clustering–based regularization strategy—thereby jointly preserving geometric priors and ensuring training stability. The method unifies hyperbolic proxy loss, Euclidean proxy loss, and hierarchical structural constraints into an end-to-end trainable objective. Evaluated on four benchmark datasets, our method achieves state-of-the-art performance, improving average Recall@1 by 3.2% and significantly enhancing optimization convergence stability.

0 citationsRead paper

Toward the Axiomatization of Intelligence: Structure, Time, and Existence

Apr 20, 2025

This paper addresses the fundamental challenge of formalizing intelligence—characterized by conceptual vagueness, polysemy, and resistance to rigorous axiomatization. We propose a novel meta-framework grounded in set theory and category theory, offering the first axiomatic definition of intelligence as “a structure within the universe of sets endowed with temporal evolution and interactive capacity.” Innovatively, we introduce the primitive notion of *activity* to formally capture intelligence’s intrinsic time-dependence; further, we construct the *temporal category* and *intelligence category*, linked by functors that abstractly represent system evolution and imitation mechanisms. Our framework unifies and distinguishes Hebbian learning, backpropagation, and biological reflexes along structural, temporal, and biologically plausible dimensions. Moreover, it naturally extends to axiomatize higher-order mental phenomena—including consciousness and emotion—thereby establishing a new foundational paradigm for artificial intelligence theory.

0 citationsRead paper

Recent publications

Latest Papers

TriView-YOLO: Early Multi-View Fusion for Ground Penetrating Radar Cavity Detection in Soft, High-Water-Content Soils

Aug 10, 2026

This study addresses the challenge of severe signal attenuation and poor void detectability in ground-penetrating radar (GPR) data acquired in high-moisture soft soils by proposing TriView-YOLO, a novel detection model based on the YOLOv12 architecture. The method introduces, for the first time, a nine-channel input comprising three orthogonal views—vertical B-scan, horizontal C-scan, and cross-sectional B-scan—and incorporates a custom TripleInputConv layer to enable early multi-view fusion, while producing detection bounding boxes exclusively from the vertical view. Trained on real-world data collected via a vehicle-mounted 3D GPR system, the model is evaluated under a rigorously designed protocol tailored to complex field conditions lacking public benchmarks. Experimental results demonstrate a mean average precision (mAP50) of 0.558 ± 0.028 on non-augmented test data, with per-image inference requiring only 3.1 ms (23.6 GFLOPs), significantly outperforming ablated variants and confirming the critical contribution of auxiliary views to detection performance.

0 citationsRead paper

Privacy-Preserving Product-Quantized Approximate Nearest Neighbor Search Framework for Large-scale Datasets via A Hybrid of Fully Homomorphic Encryption and Trusted Execution Environment

Apr 20, 2026

Existing privacy-preserving approximate nearest neighbor (PP-ANN) methods struggle to balance security and efficiency and often overlook privacy risks during database construction, leaving them vulnerable to embedding inversion and membership inference attacks. This work proposes PPPQ-ANN, the first framework to synergistically integrate fully homomorphic encryption (FHE) and trusted execution environments (TEEs) across the entire ANN pipeline—spanning both database construction and query processing. By combining product quantization with ciphertext packing optimizations, the framework substantially reduces FHE computational overhead. Evaluated on million-scale datasets, the system achieves database construction in under two hours and sequential search throughput exceeding 50 queries per second, demonstrating practical performance while providing strong privacy guarantees.

0 citationsRead paper

From Minimal Existence to Human Definition: The CES-IMU-HSG Theoretical Framework

Oct 15, 2025

This paper addresses the lack of ontological autonomy in artificial intelligence by proposing a cross-universe mathematical logical framework grounded in Descartes’ “cogito, ergo sum” (CES) as a minimal existence axiom. Methodologically, it integrates an intermediate metaverse (IMU) with a hierarchical state grid (HSG), introduces the categorical identity “definition = state”, and couples neuro-endocrine-genetic physiological systems via category theory, Institution theory, and fibration structures to enable multi-scale modeling over 0–3D neural functional fields. The core contribution is twofold: first, it formalizes human cognition—constrained by physical embodiment—as a cross-universe temporal algorithm; second, it endows machines with an intrinsic CES axiom, thereby establishing a formally rigorous, self-referential foundation for existence. This bridges philosophical ontology and AI engineering autonomy through a mathematically continuous, ontologically grounded framework.

0 citationsRead paper

Combined Hyperbolic and Euclidean Soft Triple Loss Beyond the Single Space Deep Metric Learning

Oct 07, 2025

Existing deep metric learning (DML) methods lack proxy-based supervised losses in hyperbolic space, limiting scalability to large-scale data and semantic modeling capability. To address this, we propose the first dual-space joint optimization framework that simultaneously learns in both hyperbolic and Euclidean spaces. Our approach introduces the first hyperbolic proxy loss, coupled with a novel dual-space soft triplet loss and a hyperbolic hierarchical clustering–based regularization strategy—thereby jointly preserving geometric priors and ensuring training stability. The method unifies hyperbolic proxy loss, Euclidean proxy loss, and hierarchical structural constraints into an end-to-end trainable objective. Evaluated on four benchmark datasets, our method achieves state-of-the-art performance, improving average Recall@1 by 3.2% and significantly enhancing optimization convergence stability.

0 citationsRead paper

Toward the Axiomatization of Intelligence: Structure, Time, and Existence

Apr 20, 2025

This paper addresses the fundamental challenge of formalizing intelligence—characterized by conceptual vagueness, polysemy, and resistance to rigorous axiomatization. We propose a novel meta-framework grounded in set theory and category theory, offering the first axiomatic definition of intelligence as “a structure within the universe of sets endowed with temporal evolution and interactive capacity.” Innovatively, we introduce the primitive notion of *activity* to formally capture intelligence’s intrinsic time-dependence; further, we construct the *temporal category* and *intelligence category*, linked by functors that abstractly represent system evolution and imitation mechanisms. Our framework unifies and distinguishes Hebbian learning, backpropagation, and biological reflexes along structural, temporal, and biologically plausible dimensions. Moreover, it naturally extends to axiomatize higher-order mental phenomena—including consciousness and emotion—thereby establishing a new foundational paradigm for artificial intelligence theory.

0 citationsRead paper