arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.24062cs.AI
收藏

在泛化之前建立基础:人工智能如何与人类在因果迁移中不同

Grounding Before Generalizing: How AI Differs from Humans in Causal Transfer

Department of Automation(清华大学自动化系);Institute for Artificial Intelligence(北京大学人工智能研究院);School of Psychological and Cognitive Sciences(北京大学心理与认知科学系);State Key Laboratory of General Artificial Intelligence(通用人工智能国家重点实验室);Beijing Key Laboratory of Behavior and Mental Health(北京行为与心理健康重点实验室)

Liangru Xiang, Yuxi Ma, Zhihao Cao, Yixin Zhu, Song-Chun Zhu

AI总结:

研究探讨了人工智能与人类在因果迁移中的差异,发现AI模型在缺乏环境基础映射时难以高效迁移,而人类能利用先验结构知识。文本条件中AI表现优异,但视觉信息反而降低性能,揭示AI依赖符号处理而非多模态推理。

摘要:

提取抽象因果结构并应用于新情境是人类智能的标志。尽管大型语言模型(LLMs)和视觉语言模型(VLMs)在多种推理任务中表现优异,但其交互式因果学习能力——通过序列探索诱导潜在结构并在不同上下文中迁移——仍未明确。人类学习者在极小暴露后即可完成迁移,而经典强化学习(RL)代理则失败惨重。是否最先进的AI模型具备人类般的抽象因果结构迁移机制仍是一个开放问题。使用需要序列发现共同原因(CC)和共同效应(CE)结构的OpenLock范式,本文显示模型表现出根本性的延迟或缺失迁移:即使成功模型也需初始环境特定映射——我们称之为环境基础——才能出现效率提升,而人类则能利用先前的结构知识。在纯文本条件下,模型匹配或超过了人类的发现效率。相比之下,视觉信息——在图像-only和文本-图像条件下——整体降低了而非提升了性能,揭示了AI对符号处理的广泛依赖而非整合多模态推理。模型还表现出系统性的CC/CE不对称性,这在人类中不存在,表明是启发式偏见而非方向中性的因果抽象。这些发现表明,大规模统计学习无法产生支撑人类类比推理的脱contextualized因果模式,确立了基于环境的基础迁移作为当前LLMs和VLMs的根本限制。

英文摘要:

Extracting abstract causal structures and applying them to novel situations is a hallmark of human intelligence. While Large Language Models (LLMs) and Vision Language Models (VLMs) have shown strong performance on a wide range of reasoning tasks, their capacity for interactive causal learning -- inducing latent structures through sequential exploration and transferring them across contexts -- remains uncharacterized. Human learners accomplish such transfer after minimal exposure, whereas classical Reinforcement Learning (RL) agents fail catastrophically. Whether state-of-the-art Artificial Intelligence (AI) models possess human-like mechanisms for abstract causal structure transfer is an open question. Using the OpenLock paradigm requiring sequential discovery of Common Cause (CC) and Common Effect (CE) structures, here we show that models exhibit fundamentally delayed or absent transfer: even successful models require initial environmental-specific mapping -- what we term environmental grounding -- before efficiency gains emerge, whereas humans leverage prior structural knowledge from the very first solution attempt. In the text-only condition, models matched or exceeded human discovery efficiency. In contrast, visual information -- in both the image-only and text-and-image conditions -- overall degraded rather than enhanced performance, revealing a broad reliance on symbolic processing rather than integrated multimodal reasoning. Models further exhibited systematic CC/CE asymmetries absent in humans, suggesting heuristic biases rather than direction-neutral causal abstraction. These findings reveal that large-scale statistical learning does not produce the decontextualized causal schemas underpinning human analogical reasoning, establishing grounding-dependent transfer as a fundamental limitation of current LLMs and VLMs.

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。

↑