本文通过GPT-4在Game of 24、Graph Coloring和STRIPS规划上的对比实验，发现自我批评导致性能下降，而可靠外部验证可显著提升性能。
摘要：
关于大型语言模型（LLM）的推理能力，一直存在相当大的意见分歧。尽管从乘法到简单规划等一系列反例，已经缓和了“推理可能会随规模扩大而自动涌现”的最初乐观情绪，但仍广泛存在一种信念：LLM能够以迭代方式自我批评并改进自己的解。这一信念似乎建立在“验证正确性应当比生成更容易”的假设之上——这是一个来自计算复杂性的相当经典的论点——但就LLM所做的是近似检索而言，该论点应当与它们无关。本文中，我们着手系统研究迭代提示在推理与规划情境下的有效性。我们提出了一项有原则的实证研究，考察GPT-4在三个领域中的表现：Game of 24、Graph Coloring和STRIPS规划。我们既实验了模型批评其自身答案，也实验了由外部正确推理器验证所提出的解。在每种情况下，我们都分析批评内容是否真正影响最终性能，以及能否在不损失性能的情况下消融增强系统中的元素。我们观察到，自我批评会造成显著的性能崩溃，而可靠的外部验证则带来显著的性能提升。我们还注意到，仅使用可靠验证器重新提示，就能保持更复杂设置的大部分收益。
英文摘要：
There has been considerable divergence of opinion on the reasoning abilities of Large Language Models (LLMs). While the initial optimism that reasoning might emerge automatically with scale has been tempered thanks to a slew of counterexamples--ranging from multiplication to simple planning--there persists a wide spread belief that LLMs can self-critique and improve their own solutions in an iterative fashion. This belief seemingly rests on the assumption that verification of correctness should be easier than generation--a rather classical argument from computational complexity--which should be irrelevant to LLMs to the extent that what they are doing is approximate retrieval. In this paper, we set out to systematically investigate the effectiveness of iterative prompting in the context of reasoning and planning. We present a principled empirical study of the performance of GPT-4 in three domains: Game of 24, Graph Coloring, and STRIPS planning. We experiment both with the model critiquing its own answers and with an external correct reasoner verifying proposed solutions. In each case, we analyze whether the content of criticisms actually affects bottom line performance, and whether we can ablate elements of the augmented system without losing performance. We observe significant performance collapse with self-critique and significant performance gains with sound external verification. We also note that merely re-prompting with a sound verifier maintains most of the benefits of more involved setups.
HeroBench: A Benchmark for Long-Horizon Planning and Structured Reasoning in Virtual Worlds
Artyom Sorokin（AXXX）；Lomonosov Moscow State University（罗蒙诺索夫莫斯科国立大学）；Higher School of Economics（高等经济学院）；Kurchatov Institute（库尔斯克研究所）；Independent Researcher（独立研究者）
Faithful-First Reasoning, Planning, and Acting for Multimodal LLMs
Shanghai Jiao Tong University（上海交通大学）；The Hong Kong University of Science and Technology（香港科技大学）；The Australian National University（澳大利亚国立大学）；Fudan University（复旦大学）；University of Sydney（悉尼大学）
SVSR: A Self-Verification and Self-Rectification Paradigm for Multimodal Reasoning
South China Agricultural University（华南农业大学）；University of Glasgow（格拉斯哥大学）；University of Electronic Science and Technology of China（电子科技大学）；Monash University（莫纳什大学）；University of Science and Technology of China（中国科学技术大学）；National University of Defense Technology（国防科技大学）；Renmin University of China（中国人民大学）；South China Normal University（华南师范大学）
A Closer Look at the Self-Verification Abilities of Large Language Models in Logical Reasoning
Institute for Artificial Intelligence, Tsinghua University（清华大学人工智能研究院）；Beijing National Research Center for Information Science and Technology（北京信息科学与技术国家研究中心）；Tencent AI Lab（腾讯AI实验室）
NCV: A Node-Wise Consistency Verification Approach for Low-Cost Structured Error Localization in LLM Reasoning
Beijing University of Posts and Telecommunications（北京邮电大学）；Shanghai Jiao Tong University（上海交通大学）；Ant Group（蚂蚁集团）；Tsinghua University（清华大学）；Inner Mongolia Research Institute of SJTU（内蒙古大学SJTU研究所）
Non-myopic Generation of Language Models for Reasoning and Planning
The University of Hong Kong（香港大学）；Peking University（北京大学）；Zhejiang University（浙江大学）；Westlake University（西湖大学）；The Hong Kong University of Science and Technology（香港科技大学）
PerceptTwin: Semantic Scene Reconstruction for Iterative LLM Planning and Verification
Department of Computer Science and Operations Research, Université de Montréal（蒙特利尔大学计算机科学与运筹学系）；Mila - Quebec AI Institute（魁北克人工智能研究所）；CIFAR AI Chair（CIFAR人工智能主席）
RAG-Star: Enhancing Deliberative Reasoning with Retrieval Augmented Verification and Refinement
Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学高瓴人工智能学院）；Wuhan University of Science and Technology（武汉科技大学）；National University of Singapore（新加坡国立大学）；Northeastern University at Qinhuangdao（东北大学秦皇岛分校）；BOSS Zhipin（BOSS直聘）
Forward-Backward Reasoning in Large Language Models for Mathematical Verification
Southern University of Science and Technology（南方科技大学）；Hong Kong University of Science and Technology（香港科技大学）；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）；Peking University（北京大学）
More Than Meets the Eye? Uncovering the Reasoning-Planning Disconnect in Training Vision-Language Driving Models
S-Lab, Nanyang Technological University, Singapore（南洋理工大学S实验室）；College of Computing and Data Science, Nanyang Technological University, Singapore（南洋理工大学计算与数据科学学院）
Why Reasoning Fails to Plan: A Planning-Centric Analysis of Long-Horizon Decision Making in LLM Agents
University of Notre Dame（诺丁汉大学）；Stanford University（斯坦福大学）；Purdue University（普渡大学）；University of Edinburgh（爱丁堡大学）；University of Oxford（牛津大学）；Yale University（耶鲁大学）
