循环变换器中逐步数据归因
Step-resolved data attribution for looped transformers
AI总结：
本文提出SDI方法，用于分析循环变换器中每个循环步骤对模型的影响，提升数据归因和可解释性能力。
摘要：
我们研究了单个训练示例如何塑造循环变换器的内部计算，其中共享块应用于τ次循环迭代以实现潜在推理。现有的训练数据影响估计器，如TracIn，会为所有循环迭代汇总成一个标量分数，模糊了训练示例在循环计算中何时起作用。我们引入了Step-Decomposed Influence (SDI)，通过展开循环计算图，将TracIn分解为长度为τ的影响轨迹，并将影响归因于特定的循环迭代。为了使SDI在变换器规模上实用，我们提出了一个TensorSketch实现，它从不生成每个示例的梯度。在循环GPT风格模型和算法推理任务上的实验表明，SDI扩展性极佳，与全梯度基线相比误差低，并支持广泛的数据显示归因和可解释性任务，通过每一步对潜在推理过程的洞察。
英文摘要：
We study how individual training examples shape the internal computation of looped transformers, where a shared block is applied for $τ$ recurrent iterations to enable latent reasoning. Existing training-data influence estimators such as TracIn yield a single scalar score that aggregates over all loop iterations, obscuring when during the recurrent computation a training example matters. We introduce \textit{Step-Decomposed Influence (SDI)}, which decomposes TracIn into a length-$τ$ influence trajectory by unrolling the recurrent computation graph and attributing influence to specific loop iterations. To make SDI practical at transformer scale, we propose a TensorSketch implementation that never materialises per-example gradients. Experiments on looped GPT-style models and algorithmic reasoning tasks show that SDI scales excellently, matches full-gradient baselines with low error and supports a broad range of data attribution and interpretability tasks with per-step insights into the latent reasoning process.
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。
