Landmark-Guided Policy Optimization for Multi-Objective Language Model Selection

Marcio Monteiro, Weichen Li, Puyu Wang, Marius Kloft, Sophie Fellenz
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:90019-90040, 2026.

Abstract

Selecting a pretrained large language model (LLM) to fine-tune for a task-specific dataset can be time-consuming and costly. With several candidate models available to choose from, varying in size, architecture, and pretraining data, finding the best model for a specific task often involves extensive trial and error. In addition, the "best" model may not necessarily be the one with the lowest test loss, as practical considerations such as deployment costs, inference throughput, and limited search budgets might also play crucial roles. To address this, we introduce LAMPS (LAnguage Model Pareto Selection), a novel and open-source multi-objective AutoML framework that meta-learns a resource allocation policy to efficiently identify (or approximate) the Pareto front of candidate LLMs for a task-specific dataset. It is based on two key ideas: (1) landmark fine-tuning, which generates early performance indicators of the candidate models, and (2) meta-learning via reinforcement learning, which learns an effective selection policy from historical performance data (a meta-dataset). Our results show that, on held-out datasets, LAMPS reduces search time by an average of 73% compared to exhaustive search, while still covering more than 99% of the optimal target space hypervolume.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-monteiro26a, title = {Landmark-Guided Policy Optimization for Multi-Objective Language Model Selection}, author = {Monteiro, Marcio and Li, Weichen and Wang, Puyu and Kloft, Marius and Fellenz, Sophie}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {90019--90040}, 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/monteiro26a/monteiro26a.pdf}, url = {https://proceedings.mlr.press/v306/monteiro26a.html}, abstract = {Selecting a pretrained large language model (LLM) to fine-tune for a task-specific dataset can be time-consuming and costly. With several candidate models available to choose from, varying in size, architecture, and pretraining data, finding the best model for a specific task often involves extensive trial and error. In addition, the "best" model may not necessarily be the one with the lowest test loss, as practical considerations such as deployment costs, inference throughput, and limited search budgets might also play crucial roles. To address this, we introduce LAMPS (LAnguage Model Pareto Selection), a novel and open-source multi-objective AutoML framework that meta-learns a resource allocation policy to efficiently identify (or approximate) the Pareto front of candidate LLMs for a task-specific dataset. It is based on two key ideas: (1) landmark fine-tuning, which generates early performance indicators of the candidate models, and (2) meta-learning via reinforcement learning, which learns an effective selection policy from historical performance data (a meta-dataset). Our results show that, on held-out datasets, LAMPS reduces search time by an average of 73% compared to exhaustive search, while still covering more than 99% of the optimal target space hypervolume.} }
Endnote
%0 Conference Paper %T Landmark-Guided Policy Optimization for Multi-Objective Language Model Selection %A Marcio Monteiro %A Weichen Li %A Puyu Wang %A Marius Kloft %A Sophie Fellenz %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-monteiro26a %I PMLR %P 90019--90040 %U https://proceedings.mlr.press/v306/monteiro26a.html %V 306 %X Selecting a pretrained large language model (LLM) to fine-tune for a task-specific dataset can be time-consuming and costly. With several candidate models available to choose from, varying in size, architecture, and pretraining data, finding the best model for a specific task often involves extensive trial and error. In addition, the "best" model may not necessarily be the one with the lowest test loss, as practical considerations such as deployment costs, inference throughput, and limited search budgets might also play crucial roles. To address this, we introduce LAMPS (LAnguage Model Pareto Selection), a novel and open-source multi-objective AutoML framework that meta-learns a resource allocation policy to efficiently identify (or approximate) the Pareto front of candidate LLMs for a task-specific dataset. It is based on two key ideas: (1) landmark fine-tuning, which generates early performance indicators of the candidate models, and (2) meta-learning via reinforcement learning, which learns an effective selection policy from historical performance data (a meta-dataset). Our results show that, on held-out datasets, LAMPS reduces search time by an average of 73% compared to exhaustive search, while still covering more than 99% of the optimal target space hypervolume.
APA
Monteiro, M., Li, W., Wang, P., Kloft, M. & Fellenz, S.. (2026). Landmark-Guided Policy Optimization for Multi-Objective Language Model Selection. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:90019-90040 Available from https://proceedings.mlr.press/v306/monteiro26a.html.

Related Material