一句话：提出 CIPIHunter，通过配置对测试、敏感性变异、反馈搜索与中间追踪定位深度学习框架由配置引发的预测不稳定（CIPI）。
深度学习框架支持精度模式、模型优化等多种执行配置；即使模型、输入与框架版本固定，改变配置仍可能改变预测结果，即配置引发的预测不稳定性（CIPI）。其根源是配置相关的数值扰动在计算图中传播，最终导致模型层面的有害行为。已有 API 级与模型级测试方法缺乏以配置为条件的探索和传播感知的诊断能力，难以检测 CIPI。本文提出 CIPIHunter，结合受控配置对测试、面向敏感性的变异、反馈引导搜索以及中间张量追踪，以定位可疑算子或算子组。在 13 个 PyTorch 模型、100 轮预算下，CIPIHunter 取得最高的原始 CIPI 计数和 22 个去重后的 Type I 表现，比最强基线多 83.3%，同时在总体去重表现覆盖上保持竞争力。作者共提交 29 个 PyTorch issue，开发者已确认 8 个、拒绝 5 个、标记 4 个为重复，9 个待处理，3 个未明确判定即关闭，另有 3 个被标记为高优先级，表明 CIPI 测试对框架质量保障具有实用价值。
Deep learning (DL) frameworks support diverse execution configurations, including precision modes and model optimizations. Even with a fixed model, input, and framework version, changing these configurations can change model predictions, causing Configuration-Induced Prediction Instability (CIPI). It arises when configuration-dependent numerical perturbations propagate and cause potentially harmful model-level behavior. Existing API-level and model-level testing approaches are not well suited to detecting CIPI, as these approaches lack configuration-conditioned exploration and propagation-aware diagnosis. Therefore, we present CIPIHunter, which combines controlled configuration-pair testing, sensitivity-oriented mutation, feedback-guided search, and intermediate tracing to localize suspicious operators or operator groups. On 13 PyTorch models under a 100-round budget, CIPIHunter reports the highest raw CIPI count and 22 deduplicated Type I manifestations, 83.3% more than the strongest baseline, while remaining competitive in overall deduplicated manifestation coverage. In total, we submit 29 PyTorch issue reports. Developers have acknowledged eight, rejected five, and marked four as duplicates; nine remain pending, and three have been closed without an explicit determination. Separately, three reports are labeled “high-priority”. These results demonstrate the practical utility of CIPI testing for framework quality assurance.
@inproceedings{paperbot12220,
title = {CIPIHunter: Detecting Configuration-Induced Prediction Instability in Deep Learning Frameworks},
author = {Yanzhou Mu and Shuo Meng and Mijung Kim and Xiang Chen and Chunrong Fang and Zhenyu Chen and Juan Zhai},
booktitle = {ASE},
year = {2026},
doi = {10.1145/3832783.3834345}
}