arXiv：cs.LG· Abhilash Jindal, Todd Nief, Bhanu Prakash Vangala, Shankaradithyaa V, Tvisha Malik, Anshik Sahu, Aaron Schein, Amitabh Chaudhary, Tanu Malik·· 4 小时前AI 评分32
FlexiFlow：ML 工作流中基于 Bandit 的模型切换
FlexiFlow: Bandit-based Model Switching in ML Workflows
AI 导读
FlexiFlow 是一个能在 ML 工作流中动态切换替代模型的数据流系统，当当前模型准确率偏低时自动换用更合适的模型。它采用多臂 Bandit 方法对模型排序，综合考虑模型运行时间、通过用户定义断言的概率以及工作流的计算结构，并指出标准 Thompson sampling 在此场景下不够用。实验显示，运行时切换模型并复用中间结果不仅准确率更高，相比顺序执行工作流还带来 48% 的效率提升。
正文
Abstract:Model optimizations help improve inference performance and accuracy of ML workflows. However, relying on a single model to perform inference across all data batches often fails to maximize accuracy and thus overall performance. In many cases, alternate models could perform better on specific subsets of data where a primary model underperforms. Our experiments with real ML workflows indeed show that switching models improves workflow accuracy by up to 23%. Yet, current systems lack the ability to adaptively switch between models based on performance, forcing users to manually test models in sequence. We present FlexiFlow, a dataflow system that dynamically switches between alternate models when the current model exhibits low accuracy. FlexiFlow learns to rank models using a novel multi-armed bandit approach that accounts for model runtimes, probability of passing user-defined assertions, and the computational structure of the ML workflow. We show that the standard Thompson sampling approach is insufficient for switching models in ML workflows. In contrast, our proposed approaches are effective and scales to complex real-world ML workflows. Experiments show that switching models at runtime while reusing intermediate results provides higher accuracy, but also 48% efficiency gain compared to sequential workflow runs.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.07286 [cs.LG] |
| (or arXiv:2610.07286v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.07286 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Bhanu Prakash Vangala [view email]
[v1]
Mon, 5 Oct 2026 19:20:41 UTC (531 KB)
来源：arXiv：cs.LG · arxiv.org