CARE: A Responsibility-Oriented Architecture for Domain-Scoped Resolution in Modular and Upgradeable Smart Contracts
本文提出CARE架构，通过明确责任边界和域范围的能力解析来解决模块化智能合约系统中的协调、共享状态等问题，并使用TLA+进行形式化验证。
0 citationsRead paper
Selected work
Representative Papers
Parameter-Efficient Fine-Tuning of Foundation Models for Liver Tumor Segmentation in CT
研究使用参数高效微调方法（如LoRA、QLoRA、Conv-Adapter和DiSCo）优化Segment Anything Model，以提高CT图像中肝肿瘤分割的准确性与效率。
0 citationsRead paper
Spectral Adapters for Segment Anything Model-based Segmentation of Colorectal Liver Metastases in Computed Tomography
该研究提出两种光谱适配器DiSECT和SiGA，以提高Segment Anything Model在结直肠肝转移CT图像分割中的准确性，使用少量可训练参数实现高效准确的分割。
0 citationsRead paper
Small Reasoning Models are Instruction Followers in Function Calling
研究通过引入Instruction-Followed Function Calling框架，使用小型模型在指令跟随上下文中提高大型语言模型的功能调用准确性，优于现有方法，并支持高效部署。
0 citationsRead paper
Beyond Effective Sample Size: Effective Number of Proposals for Adaptive Importance Sampling
This study addresses the limitation of traditional Effective Sample Size (ESS) in detecting proposal redundancy and collapse within Adaptive Importance Sampling. We propose Effective Number of Proposals (ENP), a novel metric integrating normalized weights with sample similarity to accurately assess non-redundant proposal contributions, thereby overcoming ESS diagnostic failures. Serving as a feedback signal, ENP effectively identifies overlooked proposal degeneracy and guides adaptive update strategies. Experimental results demonstrate that this metric significantly enhances both diagnostic accuracy and sampling quality in Population Adaptive Importance Sampling for complex distribution approximation. By providing a more reliable assessment of proposal diversity, ENP offers a robust alternative to ESS, ensuring more stable and efficient adaptation in high-dimensional inference tasks where standard diagnostics often fail to capture structural deficiencies in the proposal mixture.
0 citationsRead paper
Recent publications
Latest Papers
CARE: A Responsibility-Oriented Architecture for Domain-Scoped Resolution in Modular and Upgradeable Smart Contracts
本文提出CARE架构，通过明确责任边界和域范围的能力解析来解决模块化智能合约系统中的协调、共享状态等问题，并使用TLA+进行形式化验证。
0 citationsRead paper
Parameter-Efficient Fine-Tuning of Foundation Models for Liver Tumor Segmentation in CT
研究使用参数高效微调方法（如LoRA、QLoRA、Conv-Adapter和DiSCo）优化Segment Anything Model，以提高CT图像中肝肿瘤分割的准确性与效率。
0 citationsRead paper
Spectral Adapters for Segment Anything Model-based Segmentation of Colorectal Liver Metastases in Computed Tomography
该研究提出两种光谱适配器DiSECT和SiGA，以提高Segment Anything Model在结直肠肝转移CT图像分割中的准确性，使用少量可训练参数实现高效准确的分割。
0 citationsRead paper
Small Reasoning Models are Instruction Followers in Function Calling
研究通过引入Instruction-Followed Function Calling框架，使用小型模型在指令跟随上下文中提高大型语言模型的功能调用准确性，优于现有方法，并支持高效部署。
0 citationsRead paper
Beyond Effective Sample Size: Effective Number of Proposals for Adaptive Importance Sampling
This study addresses the limitation of traditional Effective Sample Size (ESS) in detecting proposal redundancy and collapse within Adaptive Importance Sampling. We propose Effective Number of Proposals (ENP), a novel metric integrating normalized weights with sample similarity to accurately assess non-redundant proposal contributions, thereby overcoming ESS diagnostic failures. Serving as a feedback signal, ENP effectively identifies overlooked proposal degeneracy and guides adaptive update strategies. Experimental results demonstrate that this metric significantly enhances both diagnostic accuracy and sampling quality in Population Adaptive Importance Sampling for complex distribution approximation. By providing a more reliable assessment of proposal diversity, ENP offers a robust alternative to ESS, ensuring more stable and efficient adaptation in high-dimensional inference tasks where standard diagnostics often fail to capture structural deficiencies in the proposal mixture.
0 citationsRead paper