Mixture-Greedy for Online Generative Model Selection: Is UCB Necessary in Diversity-Aware Multi-Armed Bandits?

Bahar Dibaei Nia, Farzan Farnia
Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, PMLR 337:4914-4948, 2026.

Abstract

Efficient selection among multiple generative models is increasingly important in modern generative {AI}, where sampling from suboptimal models is costly. This problem can be viewed as a *multi-armed bandit (MAB)* task. Under diversity-aware evaluation scores, a non-degenerate mixture of generators can outperform any individual model, distinguishing this MAB setting from classical best-arm identification. Prior approaches incorporate an Upper Confidence Bound ({UCB}) exploration bonus into the mixture objective. However, across multiple datasets and evaluation metrics, we observe that the {UCB} term consistently slows convergence and reduces sample efficiency. In contrast, a simple *Mixture-Greedy* strategy without explicit {UCB}-type optimism converges faster and achieves even better performance, particularly for widely used metrics such as FID and Vendi where tight confidence bounds are difficult to construct. We provide theoretical insight explaining this behavior: under structural conditions, diversity-aware objectives induce *implicit exploration* by favoring interior mixtures, leading to sampling of all arms and sublinear regret guarantees for diversity-based objectives. These results suggest that in diversity-aware multi-armed bandits, e.g., for generative model selection, exploration can arise intrinsically from the objective’s geometry. The project code is available at https://github.com/bhrdbn/Mixture-Greedy.

Cite this Paper

BibTeX
@InProceedings{pmlr-v337-nia26a, title = {Mixture-Greedy for Online Generative Model Selection: Is {UCB} Necessary in Diversity-Aware Multi-Armed Bandits?}, author = {Nia, Bahar Dibaei and Farnia, Farzan}, booktitle = {Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence}, pages = {4914--4948}, 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/nia26a/nia26a.pdf}, url = {https://proceedings.mlr.press/v337/nia26a.html}, abstract = {Efficient selection among multiple generative models is increasingly important in modern generative {AI}, where sampling from suboptimal models is costly. This problem can be viewed as a *multi-armed bandit (MAB)* task. Under diversity-aware evaluation scores, a non-degenerate mixture of generators can outperform any individual model, distinguishing this MAB setting from classical best-arm identification. Prior approaches incorporate an Upper Confidence Bound ({UCB}) exploration bonus into the mixture objective. However, across multiple datasets and evaluation metrics, we observe that the {UCB} term consistently slows convergence and reduces sample efficiency. In contrast, a simple *Mixture-Greedy* strategy without explicit {UCB}-type optimism converges faster and achieves even better performance, particularly for widely used metrics such as FID and Vendi where tight confidence bounds are difficult to construct. We provide theoretical insight explaining this behavior: under structural conditions, diversity-aware objectives induce *implicit exploration* by favoring interior mixtures, leading to sampling of all arms and sublinear regret guarantees for diversity-based objectives. These results suggest that in diversity-aware multi-armed bandits, e.g., for generative model selection, exploration can arise intrinsically from the objective’s geometry. The project code is available at https://github.com/bhrdbn/Mixture-Greedy.} }
Endnote
%0 Conference Paper %T Mixture-Greedy for Online Generative Model Selection: Is UCB Necessary in Diversity-Aware Multi-Armed Bandits? %A Bahar Dibaei Nia %A Farzan Farnia %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-nia26a %I PMLR %P 4914--4948 %U https://proceedings.mlr.press/v337/nia26a.html %V 337 %X Efficient selection among multiple generative models is increasingly important in modern generative {AI}, where sampling from suboptimal models is costly. This problem can be viewed as a *multi-armed bandit (MAB)* task. Under diversity-aware evaluation scores, a non-degenerate mixture of generators can outperform any individual model, distinguishing this MAB setting from classical best-arm identification. Prior approaches incorporate an Upper Confidence Bound ({UCB}) exploration bonus into the mixture objective. However, across multiple datasets and evaluation metrics, we observe that the {UCB} term consistently slows convergence and reduces sample efficiency. In contrast, a simple *Mixture-Greedy* strategy without explicit {UCB}-type optimism converges faster and achieves even better performance, particularly for widely used metrics such as FID and Vendi where tight confidence bounds are difficult to construct. We provide theoretical insight explaining this behavior: under structural conditions, diversity-aware objectives induce *implicit exploration* by favoring interior mixtures, leading to sampling of all arms and sublinear regret guarantees for diversity-based objectives. These results suggest that in diversity-aware multi-armed bandits, e.g., for generative model selection, exploration can arise intrinsically from the objective’s geometry. The project code is available at https://github.com/bhrdbn/Mixture-Greedy.
APA
Nia, B.D. & Farnia, F.. (2026). Mixture-Greedy for Online Generative Model Selection: Is UCB Necessary in Diversity-Aware Multi-Armed Bandits?. Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 337:4914-4948 Available from https://proceedings.mlr.press/v337/nia26a.html.

Related Material