Inclusive electron-nucleus cross section models from domain adaptation
研究使用迁移学习从碳数据预训练的深度神经网络出发，针对不同核素微调模型以改进电子-核反应截面预测。
0 citationsRead paper
Selected work
Representative Papers
Synchronizing Automata: Open Problems
该文探讨了同步自动机理论中的未解决问题，主要围绕著名的Černý猜想，讨论了避免词、状态压缩、子集同步等问题及线性代数方法。
0 citationsRead paper
Decisive Margins in Differentially Private Voting
研究通过差分隐私保护投票记录时，为保证结果准确性所需最小胜利差距，并提出相应算法及上下界分析。
0 citationsRead paper
Random Access to LZ-End: Faster and Deterministic
This work addresses the lack of efficient, deterministic data structures for random access in LZ-End compressed texts, as existing solutions either suffer from slow queries or rely on randomization and are difficult to construct efficiently. The paper presents the first deterministic and efficiently constructible LZ-End index, combining LZ-End parsing with recursive decomposition and compact indexing. It supports single-character random access in $O(\log^2(n/z))$ time and substring extraction in $O(j - i + \log^2(n/z))$ time, all within $O(z)$ space, where $z$ is the size of the LZ-End parse. The index can be built in $O(z \log^2(n/z))$ time, offering significant improvements over prior approaches in terms of determinism, construction efficiency, and query performance.
0 citationsRead paper
Metric-Guided Synthetic Image Data Rendering for Deep Learning compatible with Agentic AI
This work addresses the high cost of annotating real-world data and the lack of quantitative guidance for mitigating domain gaps between synthetic and real images in scientific vision tasks. To this end, the authors propose a programmable 3D rendering framework that systematically enhances the realism, diversity, and scale of synthetic data by incorporating quantitative metrics—such as gradient similarity and zero-shot detection performance—and encapsulates the rendering pipeline as an agent skill for automated parameter optimization. This approach represents the first integration of quantitatively guided synthetic data refinement into an agent-based framework, substantially improving model visual perception: it boosts zero-shot object detection performance and further refines small-object detection when trained on mixed real-synthetic datasets. The implementation leverages the authors’ custom Python toolkit, GraNatPy, which includes the SynthClaw agent.
0 citationsRead paper
Recent publications
Latest Papers
Inclusive electron-nucleus cross section models from domain adaptation
研究使用迁移学习从碳数据预训练的深度神经网络出发，针对不同核素微调模型以改进电子-核反应截面预测。
0 citationsRead paper
Synchronizing Automata: Open Problems
该文探讨了同步自动机理论中的未解决问题，主要围绕著名的Černý猜想，讨论了避免词、状态压缩、子集同步等问题及线性代数方法。
0 citationsRead paper
Decisive Margins in Differentially Private Voting
研究通过差分隐私保护投票记录时，为保证结果准确性所需最小胜利差距，并提出相应算法及上下界分析。
0 citationsRead paper
Random Access to LZ-End: Faster and Deterministic
This work addresses the lack of efficient, deterministic data structures for random access in LZ-End compressed texts, as existing solutions either suffer from slow queries or rely on randomization and are difficult to construct efficiently. The paper presents the first deterministic and efficiently constructible LZ-End index, combining LZ-End parsing with recursive decomposition and compact indexing. It supports single-character random access in $O(\log^2(n/z))$ time and substring extraction in $O(j - i + \log^2(n/z))$ time, all within $O(z)$ space, where $z$ is the size of the LZ-End parse. The index can be built in $O(z \log^2(n/z))$ time, offering significant improvements over prior approaches in terms of determinism, construction efficiency, and query performance.
0 citationsRead paper
Metric-Guided Synthetic Image Data Rendering for Deep Learning compatible with Agentic AI
This work addresses the high cost of annotating real-world data and the lack of quantitative guidance for mitigating domain gaps between synthetic and real images in scientific vision tasks. To this end, the authors propose a programmable 3D rendering framework that systematically enhances the realism, diversity, and scale of synthetic data by incorporating quantitative metrics—such as gradient similarity and zero-shot detection performance—and encapsulates the rendering pipeline as an agent skill for automated parameter optimization. This approach represents the first integration of quantitatively guided synthetic data refinement into an agent-based framework, substantially improving model visual perception: it boosts zero-shot object detection performance and further refines small-object detection when trained on mixed real-synthetic datasets. The implementation leverages the authors’ custom Python toolkit, GraNatPy, which includes the SynthClaw agent.
0 citationsRead paper