DomainPilot: Domain-Level Loss-Guided Two-Stage Data Mixture Optimization for Efficient Language Model Fine-Tuning

AI 中文解读

**核心亮点**：一种轻量级的数据配比优化方法，无需额外训练模型或大量筛选数据，就能让AI的性能显著提升。 **通俗解读**：训练大语言模型时，数据就像学生的教材，配比是否合理直接影响学习效果。过去专家们要么花大成本从海量数据里挑“好样本”，要么为每段数据单独打分，既耗时又容易出错。这篇论文提出的方法像是给AI装上了一个“学科进度表”——它能实时监控AI在数学、编程、常识等不同领域的学习表现，然后根据每个领域的“掌握程度”动态调整教材配比。比如发现AI的数学题正确率上不去，就自动增加数学数据的比例。整个过程只需在原训练框架里加几十行代码，像安装一个软件补丁一样简单。 **实际影响**：这意味着未来AI助手会更“聪明”且更省钱。比如你问它一道复杂的数学题或让写一段代码，它的回答会更准确，因为训练时数学和编程数据得到了针对性加强。更重要的是，企业不需要砸钱买昂贵的高质量数据筛选工具，用这种轻量方法就能让现有模型升级，最终用户用上更便宜的AI服务，体验也更好。

arXiv:2607.22769v1 Announce Type: new Abstract: The training efficacy of large language models (LLMs) is fundamentally constrained by the quality and composition of training data. Existing dynamic data scheduling methods face critical limitations in industrial-scale pretraining and supervised fine-tuning (SFT): data selection incurs prohibitive O(N) costs on terabyte-scale corpora, mixture optimization schemes introduce severe I/O bottlenecks or require training auxiliary reference models, and sample-level reweighting strategies rely on loss signals that conflate noise, difficulty, and novelty. We present DomainPilot, a domain-level loss-guided two-stage data mixture optimization framework. DomainPilot introduces token-level domain loss monitoring to capture per-domain learning dynamics during training without halting the data pipeline. Building on these signals, we propose a Scaling Law guided coarse optimization stage that fits domain-specific convergence curves and derives a principled prior for mixture adjustment. A subsequent Mixing Law guided fine optimization stage refines the mixture by modeling cross-domain interaction effects through controlled sweep experiments. The entire mechanism is realized via a patch-based architecture that injects domain-aware loss computation into existing training frameworks (e.g., MindSpeed/Megatron-LM) with only ~30 lines of framework-specific adapter code. We validate DomainPilot on the Qwen3-1.7B model during SFT. Compared to the original data mixture, our optimized mixture achieves improvements of +2% on MMLU-Redux, +1.8% on AIME24, +3.8% on LiveCodeBench v5, and +3.6% on BFCL v3, without increasing total data volume or training cost. These results demonstrate that domain-level training signals provide an effective, lightweight alternative to expensive data selection or auxiliary model training for mixture optimization.