AI & ML
Meta-RL with Bayesian Linear Task Models
arXiv cs.RO0 views
arXiv:2512.20974v5 Announce Type: replace-cross Abstract: Deep Bayesian reinforcement learning adapts to unseen tasks by inferring latent transition and reward models, but existing methods typically rely on variational posteriors and evidence lower bounds, introducing approximation error and unstable task representations. We introduce GLiBRL, a deep Bayesian RL framework that combines generalised linear task models with learnable non-linear basis functions. GLiBRL features conjugate Bayesian inference, yielding exact, sequential posterior updates over task parameters and model noise, together with a closed-form marginal likelihood that eliminates variational inference. The update is naturally permutation-invariant, allowing GLiBRL to integrate with both off- and on-policy algorithms. GLiBRL also learns task representation admitting an exact kernel identity, relating distances between task representations to kernel discrepancies over the task contexts. Compared against eight representative or recent meta reinforcement learning methods, GLiBRL achieves the highest aggregate zero-shot test performance on both the MuJoCo locomotion and MetaWorld manipulation benchmarks.
Source:https://arxiv.org/abs/2512.20974PreviousRFW-S: Everyday Deformable Objects with Independent Simulation VerificationNextCoNav-UAV: Cooperative Dual-Altitude Aerial Navigation via Stackelberg Learning
Related posts
Multimodal Masked Diffusion Autoencoder for Modeling Impaired Driving
0arXiv cs.RO
CoNav-UAV: Cooperative Dual-Altitude Aerial Navigation via Stackelberg Learning
0arXiv cs.RO
RFW-S: Everyday Deformable Objects with Independent Simulation Verification
0arXiv cs.RO
Heave-Ho: Decentralized Long-Horizon Multi-Humanoid Collaboration
0arXiv cs.RO
CriticHack: Evaluating Visual Rewards Under Robot Policy Optimization
0arXiv cs.RO
SPARSER: Sparse Variable Projection by Exploiting Separable Structure in Robotic Perception
0arXiv cs.RO