arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.23972cs.AIcs.CLcs.RO
收藏

为什么我们需要世界模型来实现通用人工智能:大语言模型失败之处以及世界模型如何可能超越

Why We Need World Models for AGI: Where LLMs Fail and How World Models May Outperform

Department of Computing Technologies(计算技术系);SRM Institute of Science and Technology(SRM科学与技术学院);Bio-Sensing and Bio-Sensors Group(生物传感与生物传感器组);Smart Automation and Communication Technologies Research Institute of Sciences and Engineering(科学与工程智能自动化与通信技术研究所);University of Sharjah, UAE(阿联酋沙迦大学);Department of Computer Engineering(计算机工程系);College of Computing and Informatics(计算与信息学院)

Feisal Alaswad, Batoul Aljaddouh, Maher Alrahhal, Poovammal E, Talal Bonny

AI总结:

本文通过提出潜在动态推理(LDI)概念和Flux环境案例研究,论证了大语言模型在因果推理、状态跟踪和长程规划上的局限性,并展示基于显式状态空间的强化学习智能体在长程游戏中显著优于纯文本LLM。

摘要:

大语言模型在语言生成和知识密集型任务中表现出色,但在需要因果推理、持久状态跟踪和长程规划的场景中仍然受限。我们认为,这些限制可能源于序列预测与对潜在环境动态进行推理之间的目标层级不匹配。为了形式化这一区别,我们引入了潜在动态推理(LDI),这是一种概念性视角,将语言和多模态观测解释为底层转移动态的部分证据。为了实证研究这一视角,我们引入了Flux,一个完全通过自然语言规则指定的序列推理环境。作为一个概念验证案例研究,这些规则首先被编译成一个显式的状态转移模拟器,说明在某些情况下,结构化的潜在转移动态可以从文本规则描述中操作性地提取出来。这使得我们能够在纯文本观测上运行的LLM与直接在提取的潜在状态空间中训练的强化学习智能体之间进行受控比较。在该案例研究中,能够显式访问潜在状态空间的智能体在长程游戏中表现出更稳定的行为,总胜率约为79%,而LLM仅为11%。定性分析进一步揭示了与不稳定的持久状态跟踪一致的失败模式,包括无效动作、状态跟踪错误和短程推理失败。Flux环境的完整实现可在https://github.com/FeisalAlaswad/FLUX-RL-Agent获取。在评估的设置中,这些结果表明,如果没有持久状态跟踪和转移建模的机制,仅凭强大的序列预测可能难以支持稳健的长程动态推理。

英文摘要:

Large language models achieve strong performance in language generation and knowledge-intensive tasks, yet remain limited in settings requiring causal reasoning, persistent state tracking, and long-horizon planning. We argue that these limitations may arise from an objective-level mismatch between sequence prediction and reasoning over latent environment dynamics. To formalize this distinction, we introduce Latent Dynamics Inference (LDI), a conceptual perspective that interprets language and multimodal observations as partial evidence of underlying transition dynamics. To empirically investigate this perspective, we introduce Flux, a sequential reasoning environment specified entirely through natural-language rules. As a proof-of-concept case study, the rules are first compiled into an explicit state-transition simulator, illustrating that structured latent transition dynamics can, in some cases, be operationally extracted from textual rule descriptions. This enables a controlled comparison between the LLMs operating purely over textual observations and reinforcement-learning agents trained directly within the extracted latent state space. Within this case study, agents operating with explicit access to the latent state space exhibit substantially more stable behavior in long-horizon gameplay, achieving an aggregate win rate of approximately 79% versus 11% for LLMs. Qualitative analysis further reveals failure modes consistent with unstable persistent state tracking, including invalid actions, state-tracking errors, and short-horizon reasoning failures. The complete implementation of the Flux environment available at https://github.com/FeisalAlaswad/FLUX-RL-Agent Within the evaluated setting, these results suggest that strong sequence prediction alone may struggle to support robust long-horizon dynamic reasoning without mechanisms for persistent state tracking and transition modeling

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑