arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2505.09855cs.LGcs.AIcs.CL
收藏

Transformer中学习模式的进化视角

An evolutionary perspective on modes of learning in Transformers

Google DeepMind（谷歌DeepMind）；Department of Psychology, Princeton University（普林斯顿大学心理学系）

Alexander Y. Ku, Thomas L. Griffiths, Stephanie C. Y. Chan

阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

研究通过控制任务环境稳定性与提示可靠性，探讨Transformer中永久参数优化与上下文学习策略的适应性差异及任务依赖的策略转换机制。

摘要：

Transformer的成功源于其通过两种互补策略提升推理能力：通过权重学习（IWL）实现模型参数的永久优化，以及通过上下文学习（ICL）利用模型激活中的上下文信息进行临时调节。进化生物学表明，环境在不同时间尺度上的可预测性决定了类似策略的优先级。遗传进化通过世代逐渐修改基因型适应稳定环境，而环境波动则促进可塑性，使单一基因型在生命周期内表达不同表型，前提是存在可靠的提示。本文在受控任务设置（正弦回归和Omniglot分类）中操作化这些维度（环境稳定性与提示可靠性），以表征其对Transformer学习的影响。研究发现稳定环境偏好IWL，通常在条件静态时表现出明显转折；而可靠提示偏好ICL，特别是在环境波动时。进一步分析学习动态揭示了任务依赖的策略转换（ICL到IWL及反之）。研究证明这些转换由（1）策略相对于环境的渐近最优性，以及（2）获取该策略的优化成本决定，后者取决于任务结构和学习者的归纳偏置。

英文摘要：

The success of Transformers lies in their ability to improve inference through two complementary strategies: the permanent refinement of model parameters via in-weight learning (IWL), and the ephemeral modulation of inferences via in-context learning (ICL), which leverages contextual information maintained in the model's activations. Evolutionary biology tells us that the predictability of the environment across timescales predicts the extent to which analogous strategies should be preferred. Genetic evolution adapts to stable environmental features by gradually modifying the genotype over generations. Conversely, environmental volatility favors plasticity, which enables a single genotype to express different traits within a lifetime, provided there are reliable cues to guide the adaptation. We operationalize these dimensions (environmental stability and cue reliability) in controlled task settings (sinusoid regression and Omniglot classification) to characterize their influence on learning in Transformers. We find that stable environments favor IWL, often exhibiting a sharp transition when conditions are static. Conversely, reliable cues favor ICL, particularly when the environment is volatile. Furthermore, an analysis of learning dynamics reveals task-dependent transitions between strategies (ICL to IWL and vice versa). We demonstrate that these transitions are governed by (1) the asymptotic optimality of the strategy with respect to the environment, and (2) the optimization cost of acquiring that strategy, which depends on the task structure and the learner's inductive bias.

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。

↑