工业级大模型维护需在有限算力下优化数据混合，提升生成质量。
LLM Post-Training as Brownfield Maintenance: An Industrial Perspective on Dataware Engineering
- 通过有限补丁更新数据混合，避免从头训练
- 数据蒸馏转化率提升2.84倍，显著增加可用监督信号
- 适合关注模型持续迭代与工程化落地的团队
工业级后训练属于遗留系统维护模式：团队继承已部署检查点，在固定算力和数据混合预算下实现特定性能提升，同时不引发其他任务退化。维护对象日益呈现数据有形（dataware）特征——行为由精心设计的后训练数据混合决定，通过受控的数据补丁更新，而非全量重训。基于一次代码生成优化实践，本文从维护者视角揭示三大核心挑战：零和数据混合设计、以产出率为关键指标、在不确定性下的端到端集成。研究表明，进展不依赖一次性技巧，而需建立编程数据有形的工程体系。案例中，提升教师蒸馏转可用训练数据的比例后，接受的监督信号增长2.84倍，且使用相同教师模型、每问题四次尝试。主评估显示，优化后的数据补丁使CodeForces pass@1提升+2.59点（pass@3 +3.11），LiveCodeBench v6 hold-out测试集上pass@1提升+6.11点（pass@3 +8.05），所有结果在16次随机评估中均显著，内部AIME与MATH回归测试保持可接受水平。
原文摘要 · Abstract (English)
Industrial post-training is a brownfield regime. Teams inherit a deployed checkpoint and must land targeted improvements under fixed compute and mixture budgets without regressing the rest. The maintained artifact is increasingly dataware: behavior governed by a curated post-training mixture, updated via bounded mixture patches rather than clean-slate retraining. From an industrial code-generation improvement effort, we offer a maintainer's perspective on why this work is hard in practice, distilling three recurring challenges, zero-sum mixture design, yield as the binding metric, and end-to-end integration under uncertainty, and arguing that progress depends less on one-off recipes than on an engineering discipline for programming dataware. In our case study, interventions that raised the conversion of teacher distillation into usable training data increased accepted supervision by 2.84 times while using the same solution teacher and four solution attempts per candidate problem. In our primary evaluation, the yield-engineered patch improved CodeForces pass@1 by +2.59 points (+3.11 pass@3) and held-out LiveCodeBench v6 pass@1 by +6.11 (+8.05 pass@3), all statistically significant across 16 stochastic evaluations of each benchmark from one fixed checkpoint per condition, with internal AIME and MATH regression suites within tolerance.
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。