Research map / Reinforcement learning / MDPs and planning
Learning Robust State Abstractions for Hidden-Parameter Block MDPs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use HiP-BMDP with a shared universal dynamics model and learned task embeddings | Multi-task RL on DMC pixel-observation tasks · DMC extrapolation tasks | average reward | Helped |
| Use HiP-BMDP instead of the PEARL baseline | PEARL on proprioceptive-state Meta-RL tasks · Cartpole-Swingup-V0 and Walker-Walk-V1 | steps to threshold reward | Helped |
| Use the task bisimulation metric loss on task embeddings (ablation) | HiP-BMDP · Unseen DMC environments | average per-step latent model error | Helped |
| Add task embeddings without the task bisimulation metric loss (ablation) | Multi-task RL · Cheetah-Run-V0 training environment | average reward | Mixed |
| Use HiP-BMDP rather than DeepMDP, Distral, PCGrad, or GradNorm | Multi-task RL · DMC training environments | average reward | Helped |
| Use HiP-BMDP rather than the other multi-task baselines | Multi-task RL with sticky observations · DMC evaluation tasks with partial observability | average reward | Helped |
Closest papers by meaning
- MetaCARD: Meta-Reinforcement Learning with Task Uncertainty Feedback via Decoupled Context-Aware Reward and Dynamics Components (AAAI 2024 · similarity 0.88)
- Provable Benefits of Multi-task RL under Non-Markovian Decision Making Processes (ICLR 2024 · similarity 0.88)
- Decoupling Meta-Reinforcement Learning with Gaussian Task Contexts and Skills (AAAI 2024 · similarity 0.88)
- MAMBA: an Effective World Model Approach for Meta-Reinforcement Learning (ICLR 2024 · similarity 0.88)
- Towards an Information Theoretic Framework of Context-Based Offline Meta-Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Towards Large-Scale In-Context Reinforcement Learning by Meta-Training in Randomized Worlds (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.