Newest papers

METTA: Memory-Enhanced Test-Time Adaptation for Meta-Reinforcement Learning

Ali Najar, Hosein Hasani, Amir Kooshan Fattah, Yasamin Rajabi, Mahdieh Soleymani Baghshah · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31656 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Meta-reforço do ponto de aprendizagem visa treinar agentes que se adaptam rapidamente a tarefas invisíveis de interação limitada, mas isso requer a representação de evidências de tarefas sobre histórias de interação longas. Agentes recorrentes clássicos comprimem a experiência em estados ocultos, enquanto agentes baseados em transformadores recuperam de trajetórias passadas em memória quadrática e custo de computação. Neste trabalho, propomos Memory-Enhanced Test-Time Adaptation (METTA) para aprendizagem de meta-reforço. Além disso, descompõe a fase de adaptação rápida em duas escalas temporais: uma memória neural causal executa adaptação de trajetória fina dentro de episódios com complexidade de sequência linear, enquanto os resumos de episódios compact

Machine translation (Meta M2M100); the English original follows.

Abstract

Meta-reinforcement learning aims to train agents that adapt quickly to unseen tasks from limited interaction, but this requires representing task evidence over long interaction histories. Classical recurrent agents compress experience into hidden states, while transformer-based agents retrieve from past trajectories at quadratic memory and computation cost. In this work, we propose Memory-Enhanced Test-Time Adaptation (METTA) for meta-reinforcement learning. It decomposes the fast adaptation phase into two temporal scales: a causal neural memory performs fine-grained within-episode adaptation with linear sequence complexity, while compact episode-level summaries support long-range cross-episode task inference without attending over the full interaction history. Moreover, equipping agents with neural memory provides sufficient expressive capacity and test-time training capability, enabling effective adaptation to unseen held-out tasks and task families at meta-test time. On unseen held-out Meta-World ML10/45 test environments, METTA achieves a success rate over $50\%$ while maintaining a low training budget and short wall-clock time. On long-context sparse-reward PointMaze tasks, it outperforms transformer-based baselines by exhibiting more structured exploration behavior. These results show that METTA provides an efficient path toward more adaptive and scalable long-context meta-RL.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Task (project management), Adaptation (eye), Artificial intelligence, Scalability, Machine learning

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange