Research map / Reinforcement learning / Visual and representation RL
Latent Variable Representation for Reinforcement Learning
Tongzheng Ren, Chenjun Xiao, Tianjun Zhang, Na Li, Zhaoran Wang, sujay sanghavi, Dale Schuurmans, Bo Dai
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 LV-Rep-C instead of the strongest listed baseline | MuJoCo control · HalfCheetah | return: 4000.7$\pm$202.1 -> 5557.6$\pm$439.5 | Improved |
| Use LV-Rep-C instead of the strongest listed baseline | MuJoCo control · Ant-ET, Hopper-ET, Walker-ET | return: 2012.7$\pm$571.3 -> \textbf{2511.8$\pm$460.0}; 1815.5$\pm$655.1 -> \textbf{2204.8$\pm$496.0}; \textbf{2216.4$\pm$678.7} | Improved |
| Use LV-Rep-C instead of the strongest listed baseline | MuJoCo control · Reacher, Pendulum, Humanoid-ET, S-Humanoid-ET | return: -4.1$\pm$0.1 -> -5.8$\pm$0.3; 177.3$\pm$1.9 -> 167.1$\pm$3.1; 1794.4$\pm$458.3 -> {1086$\pm$278.2}; 1084.3$\pm$77.0 -> 9 | Mixed |
| Use LV-Rep-C instead of the strongest listed baseline | MuJoCo control · I-Pendulum and CartPole | return: 0.0$\pm$0.0 -> \textbf{0.0$\pm$0.0}; \textbf{200.0$\pm$0.0} -> \textbf{200.7$\pm$0.2} | No clear effect |
| Use LV-Rep instead of the strongest listed baseline | DeepMind Control Suite · cheetah-run, walker-run | return: 542.0 $\pm$ 27.7 -> \textbf{639.3$\pm$24.5}; 488.5$\pm$40.2 -> \textbf{724.2$\pm$37.8} | Improved |
| Use LV-Rep instead of the strongest listed baseline | DeepMind Control Suite · humanoid-run, hopper-hop | return: 1.1$\pm$0.1 -> \textbf{11.8$\pm$6.8}; 46.1$\pm$17.3 -> \textbf{72.9$\pm$40.6} | Improved |
| Use a discrete rather than continuous latent-variable model (ablation) | LV-Rep · MuJoCo control | return: 5557.6$\pm$439.5 -> 4616.5$\pm$261.5; -5.8$\pm$0.3 -> -6.0$\pm$0.2; {1086$\pm$278.2} -> 1359.2 $\pm$198.6; 167.1$\pm$3.1 | Mixed |
Closest papers by meaning
- Latent State Marginalization as a Low-cost Approach for Improving Exploration (ICLR 2023 · similarity 0.89)
- Reinforcement Learning Under Latent Dynamics: Toward Statistical and Algorithmic Modularity (NeurIPS 2024 · similarity 0.89)
- Constrained Latent Action Policies for Model-Based Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
- Reasoning with Latent Diffusion in Offline Reinforcement Learning (ICLR 2024 · similarity 0.86)
- Probabilistic Subgoal Representations for Hierarchical Reinforcement Learning (ICML 2024 · similarity 0.86)
- Improving Unsupervised Hierarchical Representation with Reinforcement Learning (CVPR 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.
