arXiv cs.CL· Xihe Qiu (Shanghai University of Engineering Science, National University of Singapore), Yongxin Deng (Shanghai University of Engineering Science, University of Technology Sydney), Xiaoyu Tan (National University of Singapore), Zhen Fang (University of Technology Sydney), Ling Chen (University of Technology Sydney)·· 3 小时前AI 评分24
Thought-Like-Pro：用自举式 Prolog 链式思维提升大语言模型推理能力
Thought-Like-Pro: Enhancing Reasoning of Large Language Models through Self-Bootstrapped Prolog-based Chain-of-Thought
AI 导读
研究者提出 Thought-Like-Pro 学习框架，通过模仿学习让大语言模型模仿由符号 Prolog 逻辑引擎验证并翻译的链式思维过程。该框架以提示词引导但自举的方式运行，模型从指令中生成规则与语句，借助 Prolog 引擎推导结果，再将推理轨迹转为自然语言链式思维用于模仿学习。实验显示该方法大幅提升大语言模型推理能力，结合模型平均技术在分布外推任务上性能仅小幅下降，部分数据集已开源。
来源：arXiv cs.CL · arxiv.org