Unifying and Optimizing Data Values for Selection via Sequential Decision-Making

Hongliang Chi, Qiong Wu, Zhengyi Zhou, Jonathan Light, Emily Dodwell, Yao Ma
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:19459-19493, 2026.

Abstract

Data selection has emerged as a crucial downstream application of data valuation, yet the theoretical foundations for using data values in selection remain underexplored. We reformulate data selection as a sequential decision-making problem where the optimal selection sequence arises from dynamic programming, and data values can be understood as encodings of this optimal sequence. This framework unifies and reinterprets existing methods like Data Shapley through the lens of approximate dynamic programming, revealing them as myopic linear approximations to the sequential problem. We further analyze how selection optimality degrades with utility curvature under submodularity, explaining when and why these approximations fail. To bridge theory and practice, we propose an efficient bipartite graph-based surrogate that preserves submodular structure while enabling scalable greedy selection with provable guarantees. Experiments on classical ML benchmarks and large-scale LLM fine-tuning data selection demonstrate substantial improvements over existing methods. Code is publicly available at https://github.com/frankhlchi/SeqDataVal

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-chi26d, title = {Unifying and Optimizing Data Values for Selection via Sequential Decision-Making}, author = {Chi, Hongliang and Wu, Qiong and Zhou, Zhengyi and Light, Jonathan and Dodwell, Emily and Ma, Yao}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {19459--19493}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/chi26d/chi26d.pdf}, url = {https://proceedings.mlr.press/v306/chi26d.html}, abstract = {Data selection has emerged as a crucial downstream application of data valuation, yet the theoretical foundations for using data values in selection remain underexplored. We reformulate data selection as a sequential decision-making problem where the optimal selection sequence arises from dynamic programming, and data values can be understood as encodings of this optimal sequence. This framework unifies and reinterprets existing methods like Data Shapley through the lens of approximate dynamic programming, revealing them as myopic linear approximations to the sequential problem. We further analyze how selection optimality degrades with utility curvature under submodularity, explaining when and why these approximations fail. To bridge theory and practice, we propose an efficient bipartite graph-based surrogate that preserves submodular structure while enabling scalable greedy selection with provable guarantees. Experiments on classical ML benchmarks and large-scale LLM fine-tuning data selection demonstrate substantial improvements over existing methods. Code is publicly available at https://github.com/frankhlchi/SeqDataVal} }
Endnote
%0 Conference Paper %T Unifying and Optimizing Data Values for Selection via Sequential Decision-Making %A Hongliang Chi %A Qiong Wu %A Zhengyi Zhou %A Jonathan Light %A Emily Dodwell %A Yao Ma %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-chi26d %I PMLR %P 19459--19493 %U https://proceedings.mlr.press/v306/chi26d.html %V 306 %X Data selection has emerged as a crucial downstream application of data valuation, yet the theoretical foundations for using data values in selection remain underexplored. We reformulate data selection as a sequential decision-making problem where the optimal selection sequence arises from dynamic programming, and data values can be understood as encodings of this optimal sequence. This framework unifies and reinterprets existing methods like Data Shapley through the lens of approximate dynamic programming, revealing them as myopic linear approximations to the sequential problem. We further analyze how selection optimality degrades with utility curvature under submodularity, explaining when and why these approximations fail. To bridge theory and practice, we propose an efficient bipartite graph-based surrogate that preserves submodular structure while enabling scalable greedy selection with provable guarantees. Experiments on classical ML benchmarks and large-scale LLM fine-tuning data selection demonstrate substantial improvements over existing methods. Code is publicly available at https://github.com/frankhlchi/SeqDataVal
APA
Chi, H., Wu, Q., Zhou, Z., Light, J., Dodwell, E. & Ma, Y.. (2026). Unifying and Optimizing Data Values for Selection via Sequential Decision-Making. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:19459-19493 Available from https://proceedings.mlr.press/v306/chi26d.html.

Related Material