揭示Transformer学习隐含结构的分阶段动态过程
Understanding the Staged Dynamics of Transformers in Learning Latent Structure
- 通过分解任务观察模型分阶段习得不同结构成分
- 模型能稳定组合基础转换，但难从复杂例中拆解原子转换
- 发现特定层存在可塑性窗口，冻结会延迟或阻碍阶段完成
语言建模表明Transformer能从上下文中发现隐含结构，但其如何逐步获取结构各组成部分的动态机制仍不清晰，导致有人认为模型只是重混训练数据。本文在受控环境下使用Alchemy基准（Wang et al.,2021）研究隐含结构学习。训练一个小型解码器仅有的Transformer模型，分别处理三种任务：1）从部分上下文推断缺失的转换；2）组合简单规则解决多步序列；3）分解复杂多步示例以推断中间转换。通过将每项任务分解为可解释组件，我们发现模型以离散阶段学习不同结构成分。同时观察到不对称现象：模型能稳健组合基础转换，但难以分解复杂例子以发现原子转换。最后，借助因果干预，识别出特定层存在可塑性窗口，冻结这些层会显著延迟或阻止阶段完成。研究结果揭示了Transformer获取隐含结构的演化路径，提供了训练过程中能力发展的详细视图。
原文摘要 · Abstract (English)
Language modeling has shown us that transformers can discover latent structure from context, but the dynamics of how they acquire different components of that structure remain poorly understood, leading to assertions that models just remix training data. In this work, we use the Alchemy benchmark in a controlled setting (Wang et al.,2021) to investigate latent structure learning. We train a small decoder-only transformer on three task variants: 1) inferring missing transitions from partial contextual information, 2) composing simple rules to solve multi-transition sequences, and 3) decomposing complex multi-step examples to infer intermediate transitions. By factorizing each task into interpretable components, we show that the model learns the different latent structure components in discrete stages. We also observe an asymmetry: the model composes fundamental transitions robustly, but struggles to decompose complex examples to discover the atomic transitions. Finally, using causal interventions, we identify layer-specific plasticity windows during which freezing substantially delays or prevents stage completion. These findings provide insight into how a transformer model acquires latent structure, offering a detailed view of how capabilities evolve during training.
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。