论文：LLM 生成数据增强代码，比暴力搜索少用 22 倍候选评估
Semantic Synthesis over Brute-Force Search: LLM-Guided Transformation Design for Efficient Learning
做数据增强搜索的朋友值得看：论文用 LLM 写增强代码，CIFAR-10 上拿到 0.8877，比搜 6,000 个候选的暴力法还强，评估预算少 22 到 600 倍。顺带提醒你常用的单 epoch 快速筛选并不总是可靠。
arXiv 论文《Semantic Synthesis over Brute-Force Search》提出用 LLM 直接合成可执行的数据增强代码，替代在预定义增强空间中的暴力搜索。作者在统一训练协议下对比了 9 种策略，包括标准增强策略、暴力搜索、两个 LLM 的直接代码生成，以及在按准确率排序的转换对上微调后的 LLM 生成。
重训 20 epochs 后，LLM 生成的转换程序在所有数据集上领先：CIFAR-10 达到 0.8877，CIFAR-100 达到 0.5825，SVHN 达到 0.9588，超过了遍历 6,000 个候选的暴力搜索。CIFAR-10 和 SVHN 上最好的转换来自只用 270 个候选的微调生成器，CIFAR-100 上最好的来自只用 10 个候选的现成生成器，相比暴力搜索在生成步骤减少 22 倍候选评估，现成生成更是减少 600 倍。
论文还分析了增强搜索常用的低预算评估的可靠性：单 epoch 选择分数与 20 epoch 结果的相关性随数据集和架构变化，从 0.756 到接近零。结论指出基于代码 LLM 的语义合成是搜索式转换发现的高效替代方案，优化数据转换时需验证评估保真度。
Semantic Synthesis over Brute-Force Search: LLM-Guided Transformation Design for Efficient Learning
Data transformation pipelines shape the training distribution in computer vision, but designing them is a computationally expensive problem. We investigate whether large language models can enable efficient discovery by synthesizing executable transformation programs outside predefined augmentation spaces. We compare nine strategies under a unified training protocol, including standard augmentation policies, brute-force search, direct code generation with two LLMs, and performance-guided generation using LLMs fine-tuned on accuracy-ordered transformation pairs. When the best candidates are retrained for 20 epochs, LLM-generated transformation programs lead on every dataset, reaching 0.8877 accuracy on CIFAR-10, 0.5825 on CIFAR-100, and 0.9588 on SVHN, surpassing brute-force search over 6,000 candidates. The best transform on CIFAR-10 and SVHN comes from a fine-tuned generator using 270 candidates, while the best on CIFAR-100 comes from an off-the-shelf generator using only 10. This needs 22x fewer candidate evaluations in the generation step and 600x fewer for off-the-shelf generations, relative to brute-force search. Beyond transformation discovery, we analyze the reliability of the reduced-budget evaluation commonly used for augmentation search. Correlating one-epoch selection scores with 20-epoch outcomes reveals strong, weak, and absent predictive relationships depending on the dataset and architecture, with correlations ranging from 0.756 to near zero. A positive control indicates that range restriction alone is unlikely to explain the low correlations. Our results indicate that semantic synthesis with code LLMs provides an efficient alternative to search-based transformation discovery and highlight the importance of validating evaluation fidelity when optimizing data transformations.
本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中，该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线，反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括：数据增强、CIFAR-10、代码生成、LLM，读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明：本文原始线索采编自「arXiv cs.AI」（@Usha Shrestha, Radu Timofte, Dmitry Ignatov），经 AITOP 多维评分管线结构化提炼。