Our paper on efficient meta-reinforcement learning accepted at ICLR 2025
Our paper on Efficient Cross-Episode Meta-RL, introducing ECET (Efficient Cross-Episodic Transformers) was accepted at ICLR 2025. ECET presents a novel hierarchical transformer architecture for meta-RL that significantly improves generalization while maintaining efficient meta-training, demonstrated across MuJoCo, Meta-World and ManiSkill benchmarks.