尽管语言模型在许多其他任务中表现出色，但在道德推理方面仍然存在困难。特别是，MMLU（多任务语言理解）中的 Moral Scenarios 任务是包括 GPT-3 在内的许多语言模型表现最差的任务之一。在这项工作中，我们提出了一个新的提示框架 Thought Experiments，使用反事实来教导语言模型进行更好的道德推理。实验结果表明，我们的框架从模型中引出反事实问题和答案，这反过来有助于将 Moral Scenarios 任务的准确率与其他零样本基线相比提高 9-16%。有趣的是，与数学推理任务不同，零样本思维链（CoT）推理无法开箱即用，甚至与直接零样本相比将准确率降低了约 4%。我们进一步观察到，以 5 个少样本示例形式的最小人工监督，该任务的准确率可以提高到多达 80%。
英文摘要：
Language models still struggle on moral reasoning, despite their impressive performance in many other tasks. In particular, the Moral Scenarios task in MMLU (Multi-task Language Understanding) is among the worst performing tasks for many language models, including GPT-3. In this work, we propose a new prompting framework, Thought Experiments, to teach language models to do better moral reasoning using counterfactuals. Experiment results show that our framework elicits counterfactual questions and answers from the model, which in turn helps improve the accuracy on Moral Scenarios task by 9-16% compared to other zero-shot baselines. Interestingly, unlike math reasoning tasks, zero-shot Chain-of-Thought (CoT) reasoning doesn't work out of the box, and even reduces accuracy by around 4% compared to direct zero-shot. We further observed that with minimal human supervision in the form of 5 few-shot examples, the accuracy of the task can be improved to as much as 80%.
MPI（马克斯·普朗克研究所）；ETH Zürich（苏黎世联邦理工学院）；MIT（麻省理工学院）；Harvard（哈佛大学）；IIT Kharagpur（印度理工学院卡拉格普尔分校）；Carnegie Mellon University（卡内基梅隆大学）；University of Michigan（密歇根大学）；MPI for Intelligent Systems（马克斯·普朗克智能系统研究所）
AI systems are becoming increasingly intertwined with human life. In order to effectively collaborate with humans and ensure safety, AI systems need to be able to understand, interpret and predict human moral judgments and decisions. Human moral judgments are often guided by rules, but not always. A central challenge for AI safety is capturing the flexibility of the human moral mind -- the ability to determine when a rule should be broken, especially in novel or unusual situations. In this paper, we present a novel challenge set consisting of rule-breaking question answering (RBQA) of cases th…
Thought-Path Contrastive Learning via Premise-Oriented Data Augmentation for Logical Reading Comprehension
School of Computer Science and Technology, Beijing Institute of Technology（北京理工大学计算机科学与技术学院）；Beijing Engineering Research Center of High Volume Language Information Processing and Cloud Computing Applications, Beijing Institute of Technology（北京理工大学海量语言信息处理与云计算应用北京工程研究中心）
Are Language Models Consequentialist or Deontological Moral Reasoners?
Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）；University of Washington（华盛顿大学）；University of Zürich（苏黎世大学）；University of Michigan（密歇根大学）；University of Toronto（多伦多大学）；Vector Institute（向量研究所）
CLOMO: Counterfactual Logical Modification with Large Language Models
City University of Hong Kong（香港城市大学）；Tsinghua University（清华大学）；Tencent AI Lab（腾讯AI实验室）；The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；City University of Hong Kong Shenzhen Research Institute（香港城市大学深圳研究院）；Shenzhen Campus of Sun Yat-sen University（中山大学深圳校区）；MBZUAI（Mohamed bin Zayed 人工智能大学）；DarkMatter AI Research（DarkMatter AI 研究院）
Reasoning-Token Spikes Under Prompted Untruthful Responses in Large Language Models
Chapman University（查普曼大学）；INSERM U992, Cognitive Neuroimaging Unit, NeuroSpin Center（法国国家健康与医学研究院U992认知神经影像单元，神经自旋中心）；University of California, Los Angeles（加州大学洛杉矶分校）；California Institute of Technology（加州理工学院）
Do LLM Self-Explanations Help Users Predict Model Behavior? Evaluating Counterfactual Simulatability with Pragmatic Perturbations
Faculty of Computer Science, University of Vienna（计算机科学系，维也纳大学）；UniVie Doctoral School Computer Science, University of Vienna（UniVie 计算机科学博士学院，维也纳大学）；Faculty of Philological and Cultural Studies, University of Vienna（文学与文化研究系，维也纳大学）
The past decade has witnessed dramatic gains in natural language processing and an unprecedented scaling of large language models. These developments have been accelerated by the advent of few-shot techniques such as chain of thought (CoT) prompting. Specifically, CoT pushes the performance of large language models in a few-shot setup by augmenting the prompts with intermediate steps. Despite impressive results across various tasks, the reasons behind their success have not been explored. This work uses counterfactual prompting to develop a deeper understanding of CoT-based few-shot prompting …
Macro: Enhancing Multilingual Counterfactual Explanations through Alignment-as-Preference Optimization
Technische Universität Berlin（柏林技术大学）；German Research Center for Artificial Intelligence (DFKI)（德国人工智能研究中心）；University of Duisburg-Essen（杜伊斯堡- Essen大学）；LMU Munich（慕尼黑大学）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；Saarland Informatics Campus（萨尔兰州信息学校区）；BIFOLD – Berlin Institute for the Foundations of Learning and Data（柏林学习与数据基础研究院）；Centre for European Research in Trusted AI (CERTAIN)（可信人工智能欧洲研究中心）
Critical or Compliant? The Double-Edged Sword of Reasoning in Chain-of-Thought Explanations
Seoul National University（首尔国立大学）；Language Technologies Institute, Carnegie Mellon University（语言技术研究所，卡内基梅隆大学）；Human-Computer Interaction Institute, Carnegie Mellon University（人机交互研究所，卡内基梅隆大学）
Beyond One World: Benchmarking Super Heros in Role-Playing Across Multiversal Contexts
Faculty of Engineering, Thammasat School of Engineering, Thammasat University（工程学院，泰国朱拉安高校工程学院，泰国朱拉安大学）；Department of Computer Engineering and Digital Technology, Faculty of Engineering, Chulalongkorn University（计算机工程与数字技术系，工程学院，朱拉安大学）；Artificial Intelligence Association of Thailand（泰国人工智能协会）；School of Biomedical Engineering & Imaging Sciences, King’s College London（生物医学工程与成像科学学院，伦敦国王学院）；Siriraj Informatics and Data Innovation Center (SIData+), Faculty of Medicine, Siriraj Hospital, Mahidol University（Siriraj信息与数据创新中心（SIData+），医学学院，Siriraj医院，玛希道大学）
本文提出Beyond One World基准，通过30位英雄、90个正典版本及正典事件与道德困境任务，评估LLM跨多元宇宙角色一致性，并以Think-Act Matching衡量推理与行动对齐。
Enhancing Ethical Explanations of Large Language Models through Iterative Symbolic Refinement
University of Manchester（曼彻斯特大学）；Idiap Research Institute（Idiap研究所）；National Biomarker Centre, CRUK-MI, University of Manchester（曼彻斯特大学英国癌症研究中心曼彻斯特研究所国家生物标志物中心）
Looking Inward: Language Models Can Learn About Themselves by Introspection
UC San Diego（加州大学圣迭戈分校）；Stanford University（斯坦福大学）；Truthful AI；Speechmatics；Eleos AI；Anthropic；Scale AI；New York University（纽约大学）；UC Berkeley（加州大学伯克利分校）
