Learning Cascade Ranking as One Network

Yunli Wang, Zhen Zhang, Zhiqiang Wang, Zixuan Yang, Yu Li, Jian Yang, Shiyang Wen, Peng Jiang, Kun Gai
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:65828-65844, 2025.

Abstract

Cascade Ranking is a prevalent architecture in large-scale top-k selection systems like recommendation and advertising platforms. Traditional training methods focus on single-stage optimization, neglecting interactions between stages. Recent advances have introduced interaction-aware training paradigms, but still struggle to 1) align training objectives with the goal of the entire cascade ranking (i.e., end-to-end recall of ground-truth items) and 2) learn effective collaboration patterns for different stages. To address these challenges, we propose LCRON, which introduces a novel surrogate loss function derived from the lower bound probability that ground truth items are selected by cascade ranking, ensuring alignment with the overall objective of the system. According to the properties of the derived bound, we further design an auxiliary loss for each stage to drive the reduction of this bound, leading to a more robust and effective top-k selection. LCRON enables end-to-end training of the entire cascade ranking system as a unified network. Experimental results demonstrate that LCRON achieves significant improvement over existing methods on public benchmarks and industrial applications, addressing key limitations in cascade ranking training and significantly enhancing system performance.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-wang25fc, title = {Learning Cascade Ranking as One Network}, author = {Wang, Yunli and Zhang, Zhen and Wang, Zhiqiang and Yang, Zixuan and Li, Yu and Yang, Jian and Wen, Shiyang and Jiang, Peng and Gai, Kun}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {65828--65844}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/wang25fc/wang25fc.pdf}, url = {https://proceedings.mlr.press/v267/wang25fc.html}, abstract = {Cascade Ranking is a prevalent architecture in large-scale top-k selection systems like recommendation and advertising platforms. Traditional training methods focus on single-stage optimization, neglecting interactions between stages. Recent advances have introduced interaction-aware training paradigms, but still struggle to 1) align training objectives with the goal of the entire cascade ranking (i.e., end-to-end recall of ground-truth items) and 2) learn effective collaboration patterns for different stages. To address these challenges, we propose LCRON, which introduces a novel surrogate loss function derived from the lower bound probability that ground truth items are selected by cascade ranking, ensuring alignment with the overall objective of the system. According to the properties of the derived bound, we further design an auxiliary loss for each stage to drive the reduction of this bound, leading to a more robust and effective top-k selection. LCRON enables end-to-end training of the entire cascade ranking system as a unified network. Experimental results demonstrate that LCRON achieves significant improvement over existing methods on public benchmarks and industrial applications, addressing key limitations in cascade ranking training and significantly enhancing system performance.} }
Endnote
%0 Conference Paper %T Learning Cascade Ranking as One Network %A Yunli Wang %A Zhen Zhang %A Zhiqiang Wang %A Zixuan Yang %A Yu Li %A Jian Yang %A Shiyang Wen %A Peng Jiang %A Kun Gai %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-wang25fc %I PMLR %P 65828--65844 %U https://proceedings.mlr.press/v267/wang25fc.html %V 267 %X Cascade Ranking is a prevalent architecture in large-scale top-k selection systems like recommendation and advertising platforms. Traditional training methods focus on single-stage optimization, neglecting interactions between stages. Recent advances have introduced interaction-aware training paradigms, but still struggle to 1) align training objectives with the goal of the entire cascade ranking (i.e., end-to-end recall of ground-truth items) and 2) learn effective collaboration patterns for different stages. To address these challenges, we propose LCRON, which introduces a novel surrogate loss function derived from the lower bound probability that ground truth items are selected by cascade ranking, ensuring alignment with the overall objective of the system. According to the properties of the derived bound, we further design an auxiliary loss for each stage to drive the reduction of this bound, leading to a more robust and effective top-k selection. LCRON enables end-to-end training of the entire cascade ranking system as a unified network. Experimental results demonstrate that LCRON achieves significant improvement over existing methods on public benchmarks and industrial applications, addressing key limitations in cascade ranking training and significantly enhancing system performance.
APA
Wang, Y., Zhang, Z., Wang, Z., Yang, Z., Li, Y., Yang, J., Wen, S., Jiang, P. & Gai, K.. (2025). Learning Cascade Ranking as One Network. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:65828-65844 Available from https://proceedings.mlr.press/v267/wang25fc.html.

Related Material