揭示大模型如何在中间层将词符转化为可复用的逻辑结构
Unveiling the Reasoning Process of Large Language Models
- 通过注意力机制分析，发现中间层负责将词符信息重组为规则级表示
- 中间层状态占据低维流形，跨词汇的符号规则对齐度更高
- 干预中间层组件导致显著性能下降，证明其是抽象推理的关键阶段
大语言模型常进行超越表面词符的推理，但词符信息转化为抽象关系结构的内部阶段尚不明确。我们通过分析自回归推理过程中注意力头与层的信息变换，研究了这一问题。在数学与符号推理任务中，观察到分层分工：外层主要保留和传递输入特征，而中层则将其重组为更易迁移的规则级表示。该结论得到表征几何支持：中层状态位于低维流形，且在不同词汇中实现相同符号规则时表现出更强对齐性。进一步因果干预显示，依据交互准则识别的中层组件被移除后，下游影响和准确率下降远超其他区域或随机移除。结果表明，抽象推理并非均匀分布于变压器各层，而是集中在将词符信息转化为可复用关系结构的中层计算阶段。
原文摘要 · Abstract (English)
Large language models often reason beyond surface tokens, but the internal stage at which token-level information becomes abstract relational structure remains unclear. We investigate this question by analyzing how attention heads and layers transform information during autoregressive reasoning. Across mathematical and symbolic reasoning tasks, we observe a consistent layer-wise division of labor: outer layers mainly preserve and route input-related features, whereas middle layers reorganize them into more transferable rule-level representations. This interpretation is supported by representation geometry: middle-layer states occupy lower-dimensional manifolds and show stronger alignment across disjoint vocabularies that instantiate the same symbolic rules. It is further supported by causal interventions: removing middle-layer components identified by our interaction-based criterion produces substantially larger downstream changes and accuracy drops than removing components from other regions or at random. Together, these results suggest that abstract reasoning is not uniformly distributed across transformer layers, but is preferentially formed in a middle-layer computation stage that converts token-level information into reusable relational structure.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。