MAGELLAN: Metacognitive predictions of learning progress guide autotelic LLM agents in large goal spaces

Loris Gaven, Thomas Carta, Clément Romac, Cédric Colas, Sylvain Lamprier, Olivier Sigaud, Pierre-Yves Oudeyer
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:18920-18953, 2025.

Abstract

Open-ended learning agents must efficiently prioritize goals in vast possibility spaces, focusing on those that maximize learning progress (LP). When such autotelic exploration is achieved by LLM agents trained with online RL in high-dimensional and evolving goal spaces, a key challenge for LP prediction is modeling one’s own competence, a form of metacognitive monitoring. Traditional approaches either require extensive sampling or rely on brittle expert-defined goal groupings. We introduce MAGELLAN, a metacognitive framework that lets LLM agents learn to predict their competence and learning progress online. By capturing semantic relationships between goals, MAGELLAN enables sample-efficient LP estimation and dynamic adaptation to evolving goal spaces through generalization. In an interactive learning environment, we show that MAGELLAN improves LP prediction efficiency and goal prioritization, being the only method allowing the agent to fully master a large and evolving goal space. These results demonstrate how augmenting LLM agents with a metacognitive ability for LP predictions can effectively scale curriculum learning to open-ended goal spaces.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-gaven25a, title = {{MAGELLAN}: Metacognitive predictions of learning progress guide autotelic {LLM} agents in large goal spaces}, author = {Gaven, Loris and Carta, Thomas and Romac, Cl\'{e}ment and Colas, C\'{e}dric and Lamprier, Sylvain and Sigaud, Olivier and Oudeyer, Pierre-Yves}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {18920--18953}, 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/gaven25a/gaven25a.pdf}, url = {https://proceedings.mlr.press/v267/gaven25a.html}, abstract = {Open-ended learning agents must efficiently prioritize goals in vast possibility spaces, focusing on those that maximize learning progress (LP). When such autotelic exploration is achieved by LLM agents trained with online RL in high-dimensional and evolving goal spaces, a key challenge for LP prediction is modeling one’s own competence, a form of metacognitive monitoring. Traditional approaches either require extensive sampling or rely on brittle expert-defined goal groupings. We introduce MAGELLAN, a metacognitive framework that lets LLM agents learn to predict their competence and learning progress online. By capturing semantic relationships between goals, MAGELLAN enables sample-efficient LP estimation and dynamic adaptation to evolving goal spaces through generalization. In an interactive learning environment, we show that MAGELLAN improves LP prediction efficiency and goal prioritization, being the only method allowing the agent to fully master a large and evolving goal space. These results demonstrate how augmenting LLM agents with a metacognitive ability for LP predictions can effectively scale curriculum learning to open-ended goal spaces.} }
Endnote
%0 Conference Paper %T MAGELLAN: Metacognitive predictions of learning progress guide autotelic LLM agents in large goal spaces %A Loris Gaven %A Thomas Carta %A Clément Romac %A Cédric Colas %A Sylvain Lamprier %A Olivier Sigaud %A Pierre-Yves Oudeyer %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-gaven25a %I PMLR %P 18920--18953 %U https://proceedings.mlr.press/v267/gaven25a.html %V 267 %X Open-ended learning agents must efficiently prioritize goals in vast possibility spaces, focusing on those that maximize learning progress (LP). When such autotelic exploration is achieved by LLM agents trained with online RL in high-dimensional and evolving goal spaces, a key challenge for LP prediction is modeling one’s own competence, a form of metacognitive monitoring. Traditional approaches either require extensive sampling or rely on brittle expert-defined goal groupings. We introduce MAGELLAN, a metacognitive framework that lets LLM agents learn to predict their competence and learning progress online. By capturing semantic relationships between goals, MAGELLAN enables sample-efficient LP estimation and dynamic adaptation to evolving goal spaces through generalization. In an interactive learning environment, we show that MAGELLAN improves LP prediction efficiency and goal prioritization, being the only method allowing the agent to fully master a large and evolving goal space. These results demonstrate how augmenting LLM agents with a metacognitive ability for LP predictions can effectively scale curriculum learning to open-ended goal spaces.
APA
Gaven, L., Carta, T., Romac, C., Colas, C., Lamprier, S., Sigaud, O. & Oudeyer, P.. (2025). MAGELLAN: Metacognitive predictions of learning progress guide autotelic LLM agents in large goal spaces. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:18920-18953 Available from https://proceedings.mlr.press/v267/gaven25a.html.

Related Material