arXiv · Machine Learning· Hossein Sholehrasa, Mohammadhossein Amini, Doina Caragea, Remya Ampadi Ramachandran, Dan Wu, Zihang Shi, Zelin Li, Lindsey Hulbert, Eder Martinez, Xuan Xu, Majid Jaberi-Douraki·· 10 小时前AI 评分39
HARP-LoRA：面向科学表格理解的层级条件秩投影低秩适配方法
HARP-LoRA: Hierarchy-Conditioned Rank Projection for Low-Rank Adaptation in Scientific Table Understanding
AI 导读
HARP-LoRA利用行列标题层级条件化低秩投影，提升科学表格信息抽取和表格推理表现。在Llama 3.1 8B上，PK F1为57.64%、WikiTQ准确率为53.15%，较LoRA分别高1.42和6.56个百分点；Gemma 3 27B上两项指标分别为70.63%和56.87%。
来源：arXiv · Machine Learning · arxiv.org