大型语言模型在持续预训练中如何习得概念？

摘要

人类主要通过概念（如"狗"）来理解世界，这些抽象的心理表征构建了感知、推理和学习的基本框架。然而，大型语言模型在持续预训练过程中如何获取、保持及遗忘此类概念，目前仍不甚明晰。本研究系统探讨了单个概念的习得与遗忘规律，以及多概念间通过干扰与协同产生的相互作用。我们将这些行为动态与模型内部的"概念回路"（即与特定概念相关的计算子图）相关联，并引入图度量指标来刻画回路结构。研究发现：（1）概念回路能提供具有统计显著性的概念学习与遗忘信号；（2）持续预训练中概念回路呈现阶段性演化模式，早期活跃度上升后逐渐衰减并趋于稳定；（3）学习增益越大的概念在后续训练中更易出现显著遗忘；（4）语义相近概念比弱相关概念产生更强干扰；（5）概念知识存在迁移性差异，部分概念能显著促进其他概念的学习。这些发现为理解概念学习动态提供了回路层级的视角，并为设计更具可解释性与鲁棒性的概念感知训练策略奠定了理论基础。

English

Human beings primarily understand the world through concepts (e.g., dog), abstract mental representations that structure perception, reasoning, and learning. However, how large language models (LLMs) acquire, retain, and forget such concepts during continual pretraining remains poorly understood. In this work, we study how individual concepts are acquired and forgotten, as well as how multiple concepts interact through interference and synergy. We link these behavioral dynamics to LLMs' internal Concept Circuits, computational subgraphs associated with specific concepts, and incorporate Graph Metrics to characterize circuit structure. Our analysis reveals: (1) LLMs concept circuits provide a non-trivial, statistically significant signal of concept learning and forgetting; (2) Concept circuits exhibit a stage-wise temporal pattern during continual pretraining, with an early increase followed by gradual decrease and stabilization; (3) concepts with larger learning gains tend to exhibit greater forgetting under subsequent training; (4) semantically similar concepts induce stronger interference than weakly related ones; (5) conceptual knowledge differs in their transferability, with some significantly facilitating the learning of others. Together, our findings offer a circuit-level view of concept learning dynamics and inform the design of more interpretable and robust concept-aware training strategies for LLMs.

大型语言模型在持续预训练中如何习得概念？

How Do Large Language Models Learn Concepts During Continual Pre-Training?

摘要

Support