Deconstructing Stereotypes: Scope-Conditioned Generation for Effective Multilingual Counterspeech
本文提出了一种新的范围条件生成框架，通过将结构化的刻板印象特征整合到大型语言模型提示中，有效生成多语言反言论，以应对在线仇恨言论。
0 citationsRead paper
Selected work
Representative Papers
Oracle, will I ever learn? A study of prediction convergence and complementarity across link prediction models
研究通过分析不同链接预测模型的互补性，探讨了它们在知识图谱中的预测收敛性和互补性问题，并提出了一种测量方法来评估模型组合的潜在性能。
0 citationsRead paper
Marker-Delimited Codes for Short-Blocklength, High-Rate Coding over Multi-Read Edit Channels
为解决DNA存储中编辑错误导致的数据检索问题，提出了一种基于标记分隔码和外层LDPC码的级联编码方案。
0 citationsRead paper
TSPFN: A Temporal Tabular Foundation Model for Physiological Time Series Classification
为了解决医疗时间序列分类中低至中等数据量下的泛化问题，本文提出TSPFN模型，通过整合结构化时间表示和位置嵌入来捕捉时间依赖性。
0 citationsRead paper
Interactive Whole Slide Images for RL-based Tumour Segmentation
This study addresses the computational bottlenecks in Whole Slide Image (WSI) analysis caused by ultra-high resolution and tumor sparsity by proposing an end-to-end reinforcement learning framework for sequential segmentation. Departing from predefined patching strategies, this method models WSIs as a multi-resolution hierarchical interaction environment. Utilizing an Actor-Critic architecture with Proximal Policy Optimization, the agent dynamically focuses on regions of interest by integrating local observations with global thumbnail guidance. Evaluated on lung adenocarcinoma WSIs, the proposed framework achieves coarse segmentation accuracy comparable to conventional methods while reducing inference time to seconds. These results demonstrate a significant improvement in WSI analysis efficiency, offering a scalable solution for processing gigapixel pathology images without compromising diagnostic precision.
0 citationsRead paper
Recent publications
Latest Papers
Deconstructing Stereotypes: Scope-Conditioned Generation for Effective Multilingual Counterspeech
本文提出了一种新的范围条件生成框架，通过将结构化的刻板印象特征整合到大型语言模型提示中，有效生成多语言反言论，以应对在线仇恨言论。
0 citationsRead paper
Oracle, will I ever learn? A study of prediction convergence and complementarity across link prediction models
研究通过分析不同链接预测模型的互补性，探讨了它们在知识图谱中的预测收敛性和互补性问题，并提出了一种测量方法来评估模型组合的潜在性能。
0 citationsRead paper
Marker-Delimited Codes for Short-Blocklength, High-Rate Coding over Multi-Read Edit Channels
为解决DNA存储中编辑错误导致的数据检索问题，提出了一种基于标记分隔码和外层LDPC码的级联编码方案。
0 citationsRead paper
TSPFN: A Temporal Tabular Foundation Model for Physiological Time Series Classification
为了解决医疗时间序列分类中低至中等数据量下的泛化问题，本文提出TSPFN模型，通过整合结构化时间表示和位置嵌入来捕捉时间依赖性。
0 citationsRead paper
Interactive Whole Slide Images for RL-based Tumour Segmentation
This study addresses the computational bottlenecks in Whole Slide Image (WSI) analysis caused by ultra-high resolution and tumor sparsity by proposing an end-to-end reinforcement learning framework for sequential segmentation. Departing from predefined patching strategies, this method models WSIs as a multi-resolution hierarchical interaction environment. Utilizing an Actor-Critic architecture with Proximal Policy Optimization, the agent dynamically focuses on regions of interest by integrating local observations with global thumbnail guidance. Evaluated on lung adenocarcinoma WSIs, the proposed framework achieves coarse segmentation accuracy comparable to conventional methods while reducing inference time to seconds. These results demonstrate a significant improvement in WSI analysis efficiency, offering a scalable solution for processing gigapixel pathology images without compromising diagnostic precision.
0 citationsRead paper