Research map / Reinforcement learning / MDPs and planning
Blockwise Sequential Model Learning for Partially Observable Reinforcement Learning
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 |
|---|---|---|---|
| add the blockwise RNN to self-attention (ablation) | Blockwise sequential model · Sequential target-reaching task (R=15) | success rate (%): 21.4 -> 91.4 | Helped |
| add self-attention to the blockwise RNN (ablation) | Blockwise sequential model · Pendulum - Random Missing Version | average return: -467.7 -> -300.2 | Helped |
| replace contribution-based selection with pooling (ablation) | Proposed compression method · Pendulum - Random Missing Version | average return: -300.2 -> -354.1 | Hurt |
| replace contribution-based selection with top-k averaging (ablation) | Proposed compression method · Pendulum - Random Missing Version | average return: -300.2 -> -352.0 | Hurt |
| replace contribution-based selection with learned linear compression (ablation) | Proposed compression method · Pendulum - Random Missing Version | average return: -300.2 -> -346.5 | Hurt |
| replace contribution-based selection with random selection (ablation) | Proposed compression method · Pendulum - Random Missing Version | average return: -300.2 -> -349.6 | Hurt |
| vary block length L while holding k=3 (ablation) | Proposed block model · Sequential target-reaching task (R=15) | success rate (%): 32.0%, 19.6%, 57.8%, 91.4% | Mixed |
| vary the number of selected elements k while holding L=32 (ablation) | Proposed block model · Sequential target-reaching task (R=15) | success rate (%): 67.7%, 91.4%, 67.3%, 66.6% | Mixed |
Closest papers by meaning
- Provable Representation with Efficient Planning for Partially Observable Reinforcement Learning (ICML 2024 · similarity 0.88)
- Provable Partially Observable Reinforcement Learning with Privileged Information (NeurIPS 2024 · similarity 0.87)
- Sequential Monte Carlo for Policy Optimization in Continuous POMDPs (NeurIPS 2025 · similarity 0.87)
- Mitigating Partial Observability in Sequential Decision Processes via the Lambda Discrepancy (NeurIPS 2024 · similarity 0.86)
- Rethinking Transformers in Solving POMDPs (ICML 2024 · similarity 0.86)
- Model-based Reinforcement Learning for Confounded POMDPs (ICML 2024 · similarity 0.86)
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.