反向诅咒是否是一个束缚问题？从基本泛化失败中揭示变换器的局限性
Is the Reversal Curse a Binding Problem? Uncovering Limitations of Transformers from a Basic Generalization Failure
AI总结：
本文通过揭示变换器在概念绑定方面的局限性，首次突破反向诅咒并提升模型泛化能力。
摘要：
尽管具有令人印象深刻的能力，LLMs表现出一种基本的泛化失败，称为反向诅咒，其中它们难以学习可逆的事实关联。理解为什么会发生这种情况可以帮助识别当前模型的弱点，并推动其泛化和鲁棒性的发展。在本文中，我们推测LLMs中的反向诅咒是认知科学、神经科学和人工智能领域长期存在的束缚问题的一种表现。具体来说，我们假设反向诅咒的两个主要原因是变换器在概念绑定方面的限制：概念表示的一致性和纠缠。我们进行了一系列实验来支持这些假设。我们的探索导致了一种基于JEPA（联合嵌入预测架构）的模型设计，首次打破了反向诅咒，而无需通过专门的数据增强或非因果掩码来规避它，并且通过结合支持解纠缠概念表示的特殊记忆层，还可以进一步提高泛化能力。我们的研究开启了更广泛的根本挑战，即设计能够以较少的人类支架学习系统性概念绑定的模型。
英文摘要：
Despite their impressive capabilities, LLMs exhibit a basic generalization failure known as the Reversal Curse, where they struggle to learn reversible factual associations. Understanding why this occurs could help identify weaknesses in current models and advance their generalization and robustness. In this paper, we conjecture that the Reversal Curse in LLMs is a manifestation of the long-standing binding problem in cognitive science, neuroscience and AI. Specifically, we hypothesize two primary causes of the Reversal Curse stemming from transformers' limitations in conceptual binding: the inconsistency and entanglements of concept representations. We perform a series of experiments that support these conjectures. Our exploration leads to a model design based on JEPA (Joint-Embedding Predictive Architecture) that for the first time breaks the Reversal Curse without side-stepping it with specialized data augmentation or non-causal masking, and moreover, generalization could be further improved by incorporating special memory layers that support disentangled concept representations. Our research opens up the broader fundamental challenge of designing models capable of learning systematic conceptual binding with less human scaffolding.
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。
