DeSyR: A Decoupled Symbolic Recovery Framework with PINN-Guided Structure Search and Physics-Informed Coefficient Refinement
本文提出DeSyR框架，通过PINN指导的结构搜索和物理信息系数优化方法，解决从神经近似中恢复紧凑显式解的问题。
0 citationsRead paper
Selected work
Representative Papers
An Ontology-Guided Multi-Anchor Graph Retrieval Framework for Traffic Legal Liability Determination
This study addresses the challenge in existing retrieval-augmented generation (RAG) approaches for traffic legal liability determination, where modeling interdependent legal provisions across multiple legal dimensions remains difficult due to a multi-dimensional retrieval bottleneck. To overcome this limitation, the authors propose OMAGR, a novel framework that introduces, for the first time, an ontology-guided multi-anchor parallel graph retrieval mechanism. Specifically, the query is decomposed into multiple semantic anchors via ontology alignment, enabling parallel graph-based retrieval across distinct legal dimensions, followed by result fusion to enhance generation accuracy. This approach substantially mitigates information compression and omission issues inherent in joint retrieval of multi-dimensional legal provisions. Experimental results on the newly constructed TrafficLaw-QA dataset demonstrate that TrafficOmni-RAG significantly outperforms current baselines in both Context Precision and Faithfulness metrics.
0 citationsRead paper
Graph2Idea:Retrieval-Augmented Scientific Idea Generation with Graph-Structured Contexts
Existing large language model–based approaches for scientific idea generation rely on flat text retrieval, which struggles to explicitly model the relationships among problems, methods, mechanisms, and findings across papers, often resulting in redundant and weakly relevant contexts. This work proposes Graph2Idea, a novel framework that introduces knowledge graphs into retrieval-augmented scientific idea generation. By converting retrieved literature into structured triples and dynamically constructing a target-centered knowledge graph, the method extracts compact, traceable graph-structured contexts. A two-stage generation pipeline then guides the large language model to synthesize novel, feasible, and high-quality research ideas. Experimental results demonstrate that Graph2Idea significantly outperforms existing methods on a scientific idea generation benchmark, with notable improvements in novelty (0.45→0.52), quality (0.24→0.29), and feasibility (0.22→0.28).
0 citationsRead paper
TrafficRAG: A Multimodal RAG Framework for Traffic Accident Liability Determination
This study addresses the challenges of low efficiency, high subjectivity, and inconsistent outcomes in traffic accident liability determination, which are exacerbated by the limitations of existing large language models in handling video noise and insufficient legal knowledge. To overcome these issues, the authors propose TrafficRAG, a novel framework that introduces multimodal retrieval-augmented generation to this domain for the first time. TrafficRAG leverages a vision-language model to produce structured accident descriptions, retrieves relevant legal statutes and precedents through a hybrid approach combining BM25 and dense embeddings, and synthesizes multimodal evidence with legal knowledge via a large language model to generate standardized liability analysis reports. Experimental results demonstrate that the proposed method achieves 77.32% accuracy in legal applicability, 81.71% factual faithfulness, and a mean absolute error of 5.48% in liability proportion estimation, significantly outperforming baseline models.
0 citationsRead paper
EvoGens: A Population-Based Heuristic Search Framework for Scientific Idea Generation
This work addresses the tendency of current large language models to suffer from semantic convergence when generating scientific ideas, which undermines both novelty and diversity. To mitigate premature convergence, the study introduces an evolutionary computation framework—the first of its kind applied to this task—featuring a population-based search strategy. The proposed approach incorporates rank-driven mutation, semantic-aware crossover, differentiated retrieval planning, and a lightweight evaluation mechanism. Experimental results demonstrate that the method significantly outperforms baseline approaches, improving automatically assessed novelty from 0.1 to 0.4 and diversity from 0.24 to 0.55, while maintaining high idea quality.
0 citationsRead paper
Recent publications
Latest Papers
DeSyR: A Decoupled Symbolic Recovery Framework with PINN-Guided Structure Search and Physics-Informed Coefficient Refinement
本文提出DeSyR框架，通过PINN指导的结构搜索和物理信息系数优化方法，解决从神经近似中恢复紧凑显式解的问题。
0 citationsRead paper
An Ontology-Guided Multi-Anchor Graph Retrieval Framework for Traffic Legal Liability Determination
This study addresses the challenge in existing retrieval-augmented generation (RAG) approaches for traffic legal liability determination, where modeling interdependent legal provisions across multiple legal dimensions remains difficult due to a multi-dimensional retrieval bottleneck. To overcome this limitation, the authors propose OMAGR, a novel framework that introduces, for the first time, an ontology-guided multi-anchor parallel graph retrieval mechanism. Specifically, the query is decomposed into multiple semantic anchors via ontology alignment, enabling parallel graph-based retrieval across distinct legal dimensions, followed by result fusion to enhance generation accuracy. This approach substantially mitigates information compression and omission issues inherent in joint retrieval of multi-dimensional legal provisions. Experimental results on the newly constructed TrafficLaw-QA dataset demonstrate that TrafficOmni-RAG significantly outperforms current baselines in both Context Precision and Faithfulness metrics.
0 citationsRead paper
Graph2Idea:Retrieval-Augmented Scientific Idea Generation with Graph-Structured Contexts
Existing large language model–based approaches for scientific idea generation rely on flat text retrieval, which struggles to explicitly model the relationships among problems, methods, mechanisms, and findings across papers, often resulting in redundant and weakly relevant contexts. This work proposes Graph2Idea, a novel framework that introduces knowledge graphs into retrieval-augmented scientific idea generation. By converting retrieved literature into structured triples and dynamically constructing a target-centered knowledge graph, the method extracts compact, traceable graph-structured contexts. A two-stage generation pipeline then guides the large language model to synthesize novel, feasible, and high-quality research ideas. Experimental results demonstrate that Graph2Idea significantly outperforms existing methods on a scientific idea generation benchmark, with notable improvements in novelty (0.45→0.52), quality (0.24→0.29), and feasibility (0.22→0.28).
0 citationsRead paper
TrafficRAG: A Multimodal RAG Framework for Traffic Accident Liability Determination
This study addresses the challenges of low efficiency, high subjectivity, and inconsistent outcomes in traffic accident liability determination, which are exacerbated by the limitations of existing large language models in handling video noise and insufficient legal knowledge. To overcome these issues, the authors propose TrafficRAG, a novel framework that introduces multimodal retrieval-augmented generation to this domain for the first time. TrafficRAG leverages a vision-language model to produce structured accident descriptions, retrieves relevant legal statutes and precedents through a hybrid approach combining BM25 and dense embeddings, and synthesizes multimodal evidence with legal knowledge via a large language model to generate standardized liability analysis reports. Experimental results demonstrate that the proposed method achieves 77.32% accuracy in legal applicability, 81.71% factual faithfulness, and a mean absolute error of 5.48% in liability proportion estimation, significantly outperforming baseline models.
0 citationsRead paper
EvoGens: A Population-Based Heuristic Search Framework for Scientific Idea Generation
This work addresses the tendency of current large language models to suffer from semantic convergence when generating scientific ideas, which undermines both novelty and diversity. To mitigate premature convergence, the study introduces an evolutionary computation framework—the first of its kind applied to this task—featuring a population-based search strategy. The proposed approach incorporates rank-driven mutation, semantic-aware crossover, differentiated retrieval planning, and a lightweight evaluation mechanism. Experimental results demonstrate that the method significantly outperforms baseline approaches, improving automatically assessed novelty from 0.1 to 0.4 and diversity from 0.24 to 0.55, while maintaining high idea quality.
0 citationsRead paper