除了合成数据,研究团队还提出了一种新型的提示策略,称为 ETA-P (Extract then Answer - Prompting: 提取逻辑链然后回答的提示)。与标准提示策略STD-P (Standard Prompting)不同,ETA-P 首先引导模型从输入文本中提取关系图,然后再尝试回答问题。这种方法类似于Graph版本的“Chain of Thought Prompting”,帮助模型系统分解复杂问题,先整理关键信息形成推理链条,再逐步推导出结论。
[1] Can Xu, et al. 2024. Wizardlm: Empowering large language models to follow complex instructions. In ICLR.
[2] Marah Abdin, et al. 2024. Phi-3 technical report: A highly capable language model locally on your phone. arXiv preprint arXiv:2404.14219.
[3] Arnav Gudibande, et al. 2023. The false promise of imitating proprietary llms. arXiv preprint arXiv:2305.15717.
[4] Yizhong Wang, et al. 2023. How far can camels go? exploring the state of instruction tuning on open resources. In NeurIPS.
[5] Jiaming Zhou, et al. 2024. Enhancing Logical Reasoning in Large Language Models through Graph-based Synthetic Data. arXiv preprint arXiv: 2409.12437.