Research map / Reinforcement learning / MDPs and planning
Learning non-Markovian Decision-Making from State-only Sequences
Aoyang Qin, Feng Gao, Qing Li, Song-Chun Zhu, Sirui Xie
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 LanMDP with context length 3 instead of context length 1 (ablation) | LanMDP · CartPole | return: 500.00$\pm$0.00 -> 500.00$\pm$0.00 | No clear effect |
| Use LanMDP with context length 3 instead of context length 1 (ablation) | LanMDP · Reacher | return: -9.00$\pm$0.87 -> -10.91$\pm$0.73 | Hurt |
| Use LanMDP with context length 2 instead of context length 1 (ablation) | LanMDP · Swimmer | return: 41.22$\pm$2.67 -> 43.52$\pm$4.31 | Improved |
| Use LanMDP with context length 3 instead of context length 1 (ablation) | LanMDP · Hopper | return: 3045.27$\pm$240.45 -> 3051.16$\pm$111.78 | No clear effect |
| Use LanMDP with context length 2 instead of context length 1 (ablation) | LanMDP · Walker2d | return: 1753.46$\pm$193.69 -> 1811.77$\pm$369.54 | No clear effect |
| Use LanMDP instead of BC | LanMDP · CartPole | return: 474.80$\pm$18.87 -> 500.00$\pm$0.00 | Improved |
| Use LanMDP instead of BC | LanMDP · Reacher | return: -8.76$\pm$0.12 -> -9.00$\pm$0.87 | Hurt |
| Use LanMDP instead of BC | LanMDP · Swimmer | return: 38.64$\pm$1.76 -> 41.22$\pm$2.67 | Improved |
| Use LanMDP instead of BC | LanMDP · Hopper | return: 3083.32$\pm$156.61 -> 3045.27$\pm$240.45 | No clear effect |
| Use LanMDP instead of BC | LanMDP · Walker2d | return: 1839.94$\pm$376.87 -> 1753.46$\pm$193.69 | No clear effect |
| Replace posterior MCMC sampling with importance sampling (ablation) | LanMDP training · Reacher, Swimmer, Hopper, Walker2d | training time per step: 0.0108/0.0076/0.0014 -> 0.0076/0.0076/0.0014; 0.0100/0.0071/0.0014 -> 0.0071/0.0071/0.0014; 0.0268/0.0170/0.0074 -> 0.01 | Improved |
Closest papers by meaning
- Masked Generative Policy for Robotic Control (ICLR 2026 · similarity 0.89)
- State-only Imitation with Transition Dynamics Mismatch (ICLR 2020 · similarity 0.89)
- Learning Markov State Abstractions for Deep Reinforcement Learning (NeurIPS 2021 · similarity 0.89)
- Causal Imitation for Markov Decision Processes: a Partial Identification Approach (NeurIPS 2024 · similarity 0.88)
- Deep Bayesian Nonparametric Learning of Rules and Plans from Demonstrations with a Learned Automaton Prior (AAAI 2020 · similarity 0.88)
- Foundation Policies with Hilbert Representations (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.
