Pessimistic Latent Task-aware Optimization for Robust Offline Meta-Reinforcement Learning
Context-based offline meta-reinforcement learning (COMRL) has advanced almost entirely through better task representations, which separate task identity from artifacts of the behavior policy so that the latent reflects the task alone.
ProofPaper ↗
Key points
- We address this gap with Pessimistic Latent Task-aware Optimization (PLATO), which exposes the policy to off-support latents during training.
- PLATO assumes a mild manifold hypothesis: task latents lie on a low-dimensional manifold, and out-of-support tasks lie farther from the training centroid.
- A learned decoder ensemble then rolls out short trajectories at each perturbed latent, and the policy is updated on the imagined transitions under a pessimism weight that down-weights steps on which the ensemble disagrees.
- We prove that the perturbation leaves the training support by a controllable margin and that the out-of-support generalization gap is bounded by the same perturbation reach, and we give explicit conditions, verified empirically, under which the epistemic disagreement at the perturbed latent stays bounded away from zero.
Sources (1)
- [1]Pessimistic Latent Task-aware Optimization for Robust Offline Meta-Reinforcement LearningarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 08:03 AM
Context-based offline meta-reinforcement learning (COMRL) has advanced almost entirely through better task representations, which separate task identity from artifacts of the behavior policy so that the latent reflects the task alone.
We address this gap with Pessimistic Latent Task-aware Optimization (PLATO), which exposes the policy to off-support latents during training.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 11, 2026ObsSpec: Accelerating Agentic Reinforcement Learning with Observation Speculation
- Oct 10, 2026CFG-Policy: Unified Diffusion Policy Learning Framework via Classifier-free Guidance
- Oct 10, 2026Bridging Risk Approximation Gaps in Model Predictive Task Sampling via In-Context Modeling
- Oct 10, 2026Getting Space Right: Agentic Rewards for Image Generation
- Oct 10, 2026Chunk-wise Model-based Policy Optimization
- Oct 9, 2026SAT-RL: Training Language Models for Logical Reasoning with Symbolic Solvers
