AI TrackerFrontier desk · AEST
Back to news
normalaitech

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

arXiv cs.AI · 23 Oct 2026, 03:58 AEST

METTA introduces a novel approach for meta-reinforcement learning that enhances test-time adaptation. It achieves over 50% success in unseen environments while maintaining efficiency in training and exploration.

Key facts

Read aloud

METTA introduces a novel approach for meta-reinforcement learning that enhances test-time adaptation. It achieves over fifty percent success in unseen environments while maintaining efficiency in training and exploration. 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. In this work, we propose Memory-Enhanced Test-Time Adaptation METTA for meta-reinforcement learning. On unseen held-out Meta-World ML10/forty five test environments, METTA achieves a success rate over fifty percent 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.

Sources in this cluster