In this work, we argue that large language models (LLMs), though trained to predict only the next token, exhibit emergent planning behaviors: $\textbf{their hidden representations encode future outputs beyond the next token}$. Through simple probing, we demonstrate that LLM prompt representations encode global attributes of their entire responses, including $\textit{structure attributes}$ (e.g., response length, reasoning steps), $\textit{content attributes}$ (e.g., character choices in storywriting, multiple-choice answers at the end of response), and $\textit{behavior attributes}$ (e.g., answer confidence, factual consistency). In addition to identifying response planning, we explore how it scales with model size across tasks and how it evolves during generation. The findings that LLMs plan ahead for the future in their hidden representations suggest potential applications for improving transparency and generation control.
Predictable Emergent Abilities of LLMs: Proxy Tasks Are All You Need
Beijing Academy of Artificial Intelligence (BAAI)(北京人工智能研究院(BAAI));China University of Mining and Technology Beijing(中国矿业大学(北京));Peking University(北京大学)
Emergent Misalignment via In-Context Learning: Narrow in-context examples can produce broadly misaligned LLMs
AIRI;MTUCI;Algoverse AI Research;Skoltech;Yandex School of Data Analysis;MSU;ISP RAS Research Center for Trusted Artificial Intelligence;Yerevan State University
Planning, Creation, Usage: Benchmarking LLMs for Comprehensive Tool Utilization in Real-World Complex Scenarios
Harbin Institute of Technology, Shenzhen(哈尔滨工业大学(深圳));Huawei Technologies Co., Ltd(华为技术有限公司);The University of Hong Kong(香港大学);Peng Cheng Laboratory, Shenzhen(鹏城实验室)
LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions
Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Fudan University(复旦大学);University of Science and Technology of China(中国科学技术大学);Shanghai Jiao Tong University(上海交通大学)
Intrinsic Guardrails: How Semantic Geometry of Personality Interacts with Emergent Misalignment in LLMs
International Institute of Information Technology, Hyderabad(国际信息研究所,海得拉巴);Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi(Mohamed bin Zayed人工智能大学,阿布扎比)
Emergent LLM behaviors are observationally equivalent to data leakage
Christopher Barrie Assistant Professor of Sociology, New York University(纽约大学社会学助理教授);New York University(纽约大学);Petter Törnberg Institute for Logic, Language, and Computation, University of Amsterdam(阿姆斯特丹大学逻辑、语言和计算研究所);University of Amsterdam(阿姆斯特丹大学)
Anticipate & Act : Integrating LLMs and Classical Planning for Efficient Task Execution in Household Environments
Robotics Research Center, IIIT Hyderabad(IIIT海得拉巴机器人研究中心);TCS Research, Tata Consultancy Services(塔塔咨询服务公司TCS研究院);CSAIL, Massachusetts Institute of Technology(麻省理工学院计算机科学与人工智能实验室);IPAB, University of Edinburgh(爱丁堡大学IPAB)
Shaping Shared Languages: Human and Large Language Models' Inductive Biases in Emergent Communication
Leiden Institute of Advanced Computer Science, Leiden University, Netherlands(莱顿高级计算机科学研究所,莱顿大学,荷兰);School of Computing, University of Kent, United Kingdom(计算学院,肯特大学,英国);Cognitive Psychology Unit, Institute of Psychology, Leiden, the Netherlands(认知心理学单元,心理学研究所,莱顿,荷兰)
Can LLM be a Good Path Planner based on Prompt Engineering? Mitigating the Hallucination for Path Planning
College of Computer Science(计算机科学学院);Sichuan Normal University(四川师范大学);School of Information and Software Engineering(信息与软件工程学院);University of Electronic Science and Technology of China(电子科技大学)
UrbanPlanBench: A Comprehensive Urban Planning Benchmark for Evaluating Large Language Models
The Thørväld Group(Thørväld集团);Inria Paris-Rocquencourt(Inria巴黎-罗克桑court分部);Rajiv Gandhi University Doimukh(拉贾格恩迪大学多伊穆克);Tsinghua University(清华大学);Palmer Research Laboratories(帕勒默研究实验室);The Kumquat Consortium(Kumquat联盟)