EUVIMEDEUVIMED entdecken
Crossref

Discovering Mixture Skills for Unsupervised Reinforcement Learning

Nelson Ma, Junyu Xuan, Guangquan Zhang, Jie Lu

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (29) · 24262-24270

DOI: 10.1609/aaai.v40i29.39606 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Nelson Ma und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Skill discovery has emerged as a popular route for unsupervised reinforcement learning (URL), offering agents a diverse, reusable set of behaviours learned before any task-specific reward is experienced. However, existing methodologies tend to favour either categorical codes or unimodal skill priors, which simplifies training at the cost of limiting the variety of behaviours they can represent. We introduce Discovery of Mixture Skills (DiMS), a URL algorithm that learns a latent Gaussian mixture by training a Gaussian Mixture Variational Autoencoder (GMVAE) in tandem with the unsupervised policy. In DiMS, a hierarchical GMVAE simultaneously discovers clusters of skills, while an auxiliary macro-latent dynamically positions mixture components to prevent mode collapse. A joint loss term combining log-likelihood and curiosity rewards enables simultaneous updates of representation and policy while improving exploration. Experiments on the Unsupervised Reinforcement Learning Benchmark (URLB) show that DiMS consistently outperforms a wide range of state-of-the-art baselines. Ablation studies confirm that the mixture prior is critical to these gains, and that DiMS is robust to alternative exploration bonuses. Overall, our results suggest that Gaussian mixture skill priors offer a compelling foundation for future unsupervised RL.

Zitieren

APA

Ma, N., Xuan, J., Zhang, G., & Lu, J. (2026). Discovering Mixture Skills for Unsupervised Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 40(29), 24262-24270. https://doi.org/10.1609/aaai.v40i29.39606

Vancouver

Ma N, Xuan J, Zhang G, Lu J. Discovering Mixture Skills for Unsupervised Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(29):24262-24270. doi:10.1609/aaai.v40i29.39606

BibTeX
@article{Ma2026,
  author = {Ma, Nelson and Xuan, Junyu and Zhang, Guangquan and Lu, Jie},
  title = {Discovering Mixture Skills for Unsupervised Reinforcement Learning},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {29},
  pages = {24262--24270},
  doi = {10.1609/aaai.v40i29.39606}
}
Kostenlos starten →