揭示大模型学习深层语义依赖的机制，解释为何思维链有效。
Syntax vs. Semantics: How Transformers Learn Deep Dependencies
- 将学习过程视为表层统计与深层语义的竞争，发现梯度饥饿现象。
- 隐式推理受抑制，导致结构化推理以突变方式出现。
- 提出新目标函数，使思维链效果提升超2倍，适合训练推理模型。
大型语言模型展现出出色的句法流畅性，但其获取深层语义依赖的优化机制仍不清晰。本文提出一个机制框架，将学习过程建模为表层统计与深层语义之间的竞争。理论分析揭示‘梯度饥饿’现象：稀疏语义依赖的误差信号在早期优化中被主动抑制，阻碍结构化推理的学习，并使其表现为突然的相变。该框架为思维链（CoT）策略的有效性提供了机制解释——将中间推理步骤显式化为具体标记，可绕过隐式推理中的抑制机制。我们在从玩具变压器到生产级模型（Llama-3.1-8B、Qwen2.5-Coder-7B）的多个规模上验证了这一发现。基于此理论，我们提出一种拓扑对齐对比目标，显式修正梯度几何。在变量绑定任务上的实验表明，该方法相比标准交叉熵微调，性能提升超过2倍。
原文摘要 · Abstract (English)
Large Language Models demonstrate remarkable syntactic fluency, yet the optimization dynamics governing their acquisition of deep semantic dependencies remain poorly understood. We propose a mechanistic framework that models this learning process as a competition between Surface Statistics and Deep Semantics. Our theoretical analysis identifies a ``Gradient Starvation" phenomenon where the error signals for sparse semantic dependencies are actively suppressed during early optimization. This suppression impedes the learning of structural reasoning and causes its emergence to manifest as a sudden phase transition. Furthermore, this framework offers a mechanistic basis for the effectiveness of Chain-of-Thought (CoT) strategies. By externalizing intermediate reasoning steps into concrete tokens, CoT effectively bypasses the suppression regime inherent to implicit reasoning. We validate these findings across scales ranging from toy transformers to production models (Llama-3.1-8B, Qwen2.5-Coder-7B). Finally, guided by this theory, we propose a topology-aligned contrastive objective that explicitly rectifies the gradient geometry. Experiments on variable binding tasks demonstrate that our method achieves an improvement that is over 2x larger than that obtained via standard cross-entropy fine-tuning.
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。