![[2509.13648] Sequential Data Augmentation for Generative Recommendation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2509.13648] Sequential Data Augmentation for Generative Recommendation
Summary
This article introduces GenPAS, a novel framework for data augmentation in generative recommendation systems, emphasizing its impact on model performance and generalization.
Why It Matters
As generative recommendation systems become increasingly integral to personalized user experiences, understanding the role of data augmentation is crucial. This research provides a systematic approach to improve model training, which can lead to better user engagement and satisfaction in various applications.
Key Takeaways
- GenPAS offers a principled framework for data augmentation in generative recommendation systems.
- Different augmentation strategies can significantly affect model performance and generalization.
- The framework allows for flexible control of training distributions, enhancing data efficiency.
Computer Science > Machine Learning arXiv:2509.13648 (cs) [Submitted on 17 Sep 2025 (v1), last revised 22 Feb 2026 (this version, v2)] Title:Sequential Data Augmentation for Generative Recommendation Authors:Geon Lee, Bhuvesh Kumar, Clark Mingxuan Ju, Tong Zhao, Kijung Shin, Neil Shah, Liam Collins View a PDF of the paper titled Sequential Data Augmentation for Generative Recommendation, by Geon Lee and 6 other authors View PDF HTML (experimental) Abstract:Generative recommendation plays a crucial role in personalized systems, predicting users' future interactions from their historical behavior sequences. A critical yet underexplored factor in training these models is data augmentation, the process of constructing training data from user interaction histories. By shaping the training distribution, data augmentation directly and often substantially affects model generalization and performance. Nevertheless, in much of the existing work, this process is simplified, applied inconsistently, or treated as a minor design choice, without a systematic and principled understanding of its effects. Motivated by our empirical finding that different augmentation strategies can yield large performance disparities, we conduct an in-depth analysis of how they reshape training distributions and influence alignment with future targets and generalization to unseen inputs. To systematize this design space, we propose GenPAS, a generalized and principled framework that models augmentation as a...
