A self-supervised auxiliary loss combining weak and strong augmentations, an adversarial discriminator, and inverse-then-forward latent dynamics improves both data efficiency and zero-shot generalization in vision-based RL.
Generalization in reinforcement learning by soft data augmentation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Self-Predictive Dynamics for Generalization of Vision-based Reinforcement Learning
A self-supervised auxiliary loss combining weak and strong augmentations, an adversarial discriminator, and inverse-then-forward latent dynamics improves both data efficiency and zero-shot generalization in vision-based RL.