claude-scholar

一位AI研究员使用claude-scholar辅助完成一篇关于大语言模型推理能力的文献综述，并生成投稿给NeurIPS的论文初稿。

我正在写一篇关于大语言模型推理能力的文献综述，目标投稿NeurIPS。请帮我：1. 搜索近3年相关高引论文；2. 总结主要研究方向（如链式思维、自我一致性等）；3. 生成论文大纲；4. 撰写引言部分。

## 文献综述：大语言模型推理能力 ### 1. 搜索论文（近3年高引） - Wei et al. (2022) "Chain-of-Thought Prompting Elicits Reasoning in Large Language Models" (NeurIPS) - 引用>5000 - Wang et al. (2023) "Self-Consistency Improves Chain of Thought Reasoning in Language Models" (ICLR) - 引用>2000 - Kojima et al. (2022) "Large Language Models are Zero-Shot Reasoners" (NeurIPS) - 引用>3000 - Yao et al. (2023) "Tree of Thoughts: Deliberate Problem Solving with Large Language Models" (NeurIPS) - 引用>1500 - Lightman et al. (2023) "Let's Verify Step by Step" (ICLR) - 引用>800 ### 2. 主要研究方向 - **链式思维提示**：通过中间推理步骤提升复杂问题解决能力 - **自我一致性**：采样多条推理路径并投票 - **思维树**：探索多条推理路径并回溯 - **过程监督**：验证每个推理步骤的正确性 - **零样本推理**：无需示例即可推理 ### 3. 论文大纲 1. Introduction 2. Background: Reasoning in LLMs 3. Chain-of-Thought Prompting 4. Self-Consistency and Ensemble Methods 5. Tree-of-Thoughts and Search-Based Reasoning 6. Process Supervision and Verification 7. Zero-Shot Reasoning 8. Challenges and Future Directions 9. Conclusion ### 4. 引言草稿 Large language model