REALITrees: Rashomon Ensemble Active Learning for Interpretable Trees

Simon Dovan Nguyen, Hayden McTavish, Kentaro Hoffman, Tyler McCormick, Cynthia Rudin
Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, PMLR 337:4832-4854, 2026.

Abstract

Active learning reduces labeling costs by querying samples to maximize information gain. A dominant framework, Query-by-Committee, typically relies on *perturbation-based diversity* by inducing model disagreement through random feature subsetting or data blinding. While this approximates one notion of epistemic uncertainty, it sacrifices direct characterization of the plausible hypothesis space. We propose the complementary approach: *{Rashomon} Ensembled Active Learning (REAL)* which constructs a committee by exhaustively enumerating the {Rashomon} Set of all near-optimal models. To address functional redundancy within this set, we adopt a Probably Approximately Correct-{Bayesian} framework using a {Gibbs} posterior to weigh committee members by their empirical risk. Leveraging recent algorithmic advances, we exactly enumerate this set for the class of sparse decision trees. Across synthetic and established active learning baselines, REAL outperforms randomized ensembles, particularly in moderately noisy environments where it leverages expanded model multiplicity to achieve faster convergence.

Cite this Paper

BibTeX
@InProceedings{pmlr-v337-nguyen26a, title = {{REALITrees}: {Rashomon} Ensemble Active Learning for Interpretable Trees}, author = {Nguyen, Simon Dovan and McTavish, Hayden and Hoffman, Kentaro and McCormick, Tyler and Rudin, Cynthia}, booktitle = {Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence}, pages = {4832--4854}, year = {2026}, editor = {Perković, Emilija and Malinsky, Daniel}, volume = {337}, series = {Proceedings of Machine Learning Research}, month = {17--21 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v337/main/assets/nguyen26a/nguyen26a.pdf}, url = {https://proceedings.mlr.press/v337/nguyen26a.html}, abstract = {Active learning reduces labeling costs by querying samples to maximize information gain. A dominant framework, Query-by-Committee, typically relies on *perturbation-based diversity* by inducing model disagreement through random feature subsetting or data blinding. While this approximates one notion of epistemic uncertainty, it sacrifices direct characterization of the plausible hypothesis space. We propose the complementary approach: *{Rashomon} Ensembled Active Learning (REAL)* which constructs a committee by exhaustively enumerating the {Rashomon} Set of all near-optimal models. To address functional redundancy within this set, we adopt a Probably Approximately Correct-{Bayesian} framework using a {Gibbs} posterior to weigh committee members by their empirical risk. Leveraging recent algorithmic advances, we exactly enumerate this set for the class of sparse decision trees. Across synthetic and established active learning baselines, REAL outperforms randomized ensembles, particularly in moderately noisy environments where it leverages expanded model multiplicity to achieve faster convergence.} }
Endnote
%0 Conference Paper %T REALITrees: Rashomon Ensemble Active Learning for Interpretable Trees %A Simon Dovan Nguyen %A Hayden McTavish %A Kentaro Hoffman %A Tyler McCormick %A Cynthia Rudin %B Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Emilija Perković %E Daniel Malinsky %F pmlr-v337-nguyen26a %I PMLR %P 4832--4854 %U https://proceedings.mlr.press/v337/nguyen26a.html %V 337 %X Active learning reduces labeling costs by querying samples to maximize information gain. A dominant framework, Query-by-Committee, typically relies on *perturbation-based diversity* by inducing model disagreement through random feature subsetting or data blinding. While this approximates one notion of epistemic uncertainty, it sacrifices direct characterization of the plausible hypothesis space. We propose the complementary approach: *{Rashomon} Ensembled Active Learning (REAL)* which constructs a committee by exhaustively enumerating the {Rashomon} Set of all near-optimal models. To address functional redundancy within this set, we adopt a Probably Approximately Correct-{Bayesian} framework using a {Gibbs} posterior to weigh committee members by their empirical risk. Leveraging recent algorithmic advances, we exactly enumerate this set for the class of sparse decision trees. Across synthetic and established active learning baselines, REAL outperforms randomized ensembles, particularly in moderately noisy environments where it leverages expanded model multiplicity to achieve faster convergence.
APA
Nguyen, S.D., McTavish, H., Hoffman, K., McCormick, T. & Rudin, C.. (2026). REALITrees: Rashomon Ensemble Active Learning for Interpretable Trees. Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 337:4832-4854 Available from https://proceedings.mlr.press/v337/nguyen26a.html.

Related Material