A hybrid meta-RL method uses offline-trained conditional neural processes to generate extra rollouts, enabling reward-free adaptation to an unseen task from a single real rollout.
A survey on offline reinforcement learning: Taxonomy, review, and open problems,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Unsupervised Meta-Testing with Conditional Neural Processes for Hybrid Meta-Reinforcement Learning
A hybrid meta-RL method uses offline-trained conditional neural processes to generate extra rollouts, enabling reward-free adaptation to an unseen task from a single real rollout.