LLMs Interpret, Embeddings Organize, Graphs Emerge: Agent-Driven Compilation of Scientific Knowledge
该研究通过使用大语言模型解释、嵌入组织和图谱构建的方法，解决了科学知识编译问题，实现了一个基于代理的科学知识系统(ASKS)。
0 citationsRead paper
Selected work
Representative Papers
RVANNS: Mixed-Precision Indexing and Locality-Aware Graph Traversal on RISC-V
This work addresses the high memory overhead and poor cache locality that hinder CPU-based approximate nearest neighbor search (ANNS). Targeting the RISC-V Vector extension (RVV), it proposes a hybrid-precision multilayer index combined with an ROrder graph node reordering strategy, which integrates vector reconstruction and distance computation to transform irregular memory accesses into forward-dense streams. The approach employs 8-bit affine bases with sparse FP16/FP32 residual representations, adjacency list remapping, and LMUL-based register grouping optimizations. Evaluated on 128-bit and 256-bit RVV processors, the design achieves speedups of 3.39× and 4.94×, respectively, delivering 2.27–2.76× higher throughput than RVV SIMD with FP32 precision and demonstrating 1.82–2.27× better energy efficiency compared to AVX-512, SVE, and GPU baselines.
0 citationsRead paper
Active Visual Perception: Opportunities and Challenges
To address the challenges of real-time visual perception and decision-making in complex dynamic environments, this work proposes an active visual perception framework that transcends traditional passive vision paradigms by enabling sensor actuation and attention-driven control to close the perception–action loop. Methodologically, it integrates computer vision, deep reinforcement learning, multimodal sensor fusion, and a lightweight real-time decision module into an end-to-end trainable active perception system. Key contributions include: (1) an online attention-guidance policy conditioned on environmental feedback; (2) a tightly coupled spatiotemporal alignment mechanism for heterogeneous multi-sensor data; and (3) a low-latency closed-loop control architecture. Experiments on robotic navigation, autonomous driving simulation, and interactive tasks demonstrate a 37% improvement in perception efficiency and a 52% reduction in decision latency, significantly enhancing adaptability and robustness in dynamic scenarios.
0 citationsRead paper
A multi-strategy improved snake optimizer for three-dimensional UAV path planning and engineering problems
To address the slow convergence and premature local optima trapping of the Snake Optimizer (SO), as well as the challenges in modeling and optimizing three-dimensional (3D) unmanned aerial vehicle (UAV) path planning, this paper proposes a Multi-Strategy enhanced Snake Optimizer (MSO). MSO integrates three novel components: (1) a sine-based adaptive random perturbation to enhance population diversity; (2) a scale-factor-modulated Lévy flight combined with a male-leader mechanism to strengthen global exploration; and (3) an elite-guided Brownian motion for position updating to accelerate local exploitation. Comprehensive evaluations on the CEC2017 (30 functions) and CEC2022 benchmark suites demonstrate that MSO significantly outperforms 11 state-of-the-art algorithms. Furthermore, MSO achieves superior accuracy, robustness, and practical applicability in real-world 3D UAV path planning and six engineering design optimization problems.
0 citationsRead paper
On the $N$th $2$-adic complexity of binary sequences identified with algebraic $2$-adic integers
This paper investigates the $N$-th order 2-adic complexity of binary sequences, focusing on its relationship with algebraic 2-adic integers. Using tools from 2-adic analysis, algebraic number theory, formal power series, and automata theory, the authors establish—for the first time—a sharp asymptotic lower bound: if the generating function $G_S(2)$ is an algebraic number of degree $d$, then the $N$-th order 2-adic complexity satisfies $ge N/d + O(1)$; the case $d = 2$ is fully characterized structurally. They further prove that 2-adic algebraic sequences coincide with automatic sequences if and only if they are ultimately periodic. Experimental evaluation confirms that such sequences exhibit high linear complexity and other desirable cryptographic properties. The results reveal a fundamental constraint imposed by algebraicity on pseudorandomness—namely, that algebraic structure inherently limits 2-adic complexity growth, thereby bounding resistance against certain algebraic attacks.
0 citationsRead paper
Recent publications
Latest Papers
LLMs Interpret, Embeddings Organize, Graphs Emerge: Agent-Driven Compilation of Scientific Knowledge
该研究通过使用大语言模型解释、嵌入组织和图谱构建的方法，解决了科学知识编译问题，实现了一个基于代理的科学知识系统(ASKS)。
0 citationsRead paper
RVANNS: Mixed-Precision Indexing and Locality-Aware Graph Traversal on RISC-V
This work addresses the high memory overhead and poor cache locality that hinder CPU-based approximate nearest neighbor search (ANNS). Targeting the RISC-V Vector extension (RVV), it proposes a hybrid-precision multilayer index combined with an ROrder graph node reordering strategy, which integrates vector reconstruction and distance computation to transform irregular memory accesses into forward-dense streams. The approach employs 8-bit affine bases with sparse FP16/FP32 residual representations, adjacency list remapping, and LMUL-based register grouping optimizations. Evaluated on 128-bit and 256-bit RVV processors, the design achieves speedups of 3.39× and 4.94×, respectively, delivering 2.27–2.76× higher throughput than RVV SIMD with FP32 precision and demonstrating 1.82–2.27× better energy efficiency compared to AVX-512, SVE, and GPU baselines.
0 citationsRead paper
Active Visual Perception: Opportunities and Challenges
To address the challenges of real-time visual perception and decision-making in complex dynamic environments, this work proposes an active visual perception framework that transcends traditional passive vision paradigms by enabling sensor actuation and attention-driven control to close the perception–action loop. Methodologically, it integrates computer vision, deep reinforcement learning, multimodal sensor fusion, and a lightweight real-time decision module into an end-to-end trainable active perception system. Key contributions include: (1) an online attention-guidance policy conditioned on environmental feedback; (2) a tightly coupled spatiotemporal alignment mechanism for heterogeneous multi-sensor data; and (3) a low-latency closed-loop control architecture. Experiments on robotic navigation, autonomous driving simulation, and interactive tasks demonstrate a 37% improvement in perception efficiency and a 52% reduction in decision latency, significantly enhancing adaptability and robustness in dynamic scenarios.
0 citationsRead paper
A multi-strategy improved snake optimizer for three-dimensional UAV path planning and engineering problems
To address the slow convergence and premature local optima trapping of the Snake Optimizer (SO), as well as the challenges in modeling and optimizing three-dimensional (3D) unmanned aerial vehicle (UAV) path planning, this paper proposes a Multi-Strategy enhanced Snake Optimizer (MSO). MSO integrates three novel components: (1) a sine-based adaptive random perturbation to enhance population diversity; (2) a scale-factor-modulated Lévy flight combined with a male-leader mechanism to strengthen global exploration; and (3) an elite-guided Brownian motion for position updating to accelerate local exploitation. Comprehensive evaluations on the CEC2017 (30 functions) and CEC2022 benchmark suites demonstrate that MSO significantly outperforms 11 state-of-the-art algorithms. Furthermore, MSO achieves superior accuracy, robustness, and practical applicability in real-world 3D UAV path planning and six engineering design optimization problems.
0 citationsRead paper
On the $N$th $2$-adic complexity of binary sequences identified with algebraic $2$-adic integers
This paper investigates the $N$-th order 2-adic complexity of binary sequences, focusing on its relationship with algebraic 2-adic integers. Using tools from 2-adic analysis, algebraic number theory, formal power series, and automata theory, the authors establish—for the first time—a sharp asymptotic lower bound: if the generating function $G_S(2)$ is an algebraic number of degree $d$, then the $N$-th order 2-adic complexity satisfies $ge N/d + O(1)$; the case $d = 2$ is fully characterized structurally. They further prove that 2-adic algebraic sequences coincide with automatic sequences if and only if they are ultimately periodic. Experimental evaluation confirms that such sequences exhibit high linear complexity and other desirable cryptographic properties. The results reveal a fundamental constraint imposed by algebraicity on pseudorandomness—namely, that algebraic structure inherently limits 2-adic complexity growth, thereby bounding resistance against certain algebraic attacks.
0 citationsRead paper