Research map / Reinforcement learning / Visual and representation RL
Sparse Graphical Memory for Robust Planning
Scott Emmons, Ajay N. Jain, Misha Laskin, Thanard Kurutach, Pieter Abbeel, Deepak Pathak
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 |
|---|---|---|---|
| Add SGM to SoRB | SoRB · PointEnv | success rate: 28.0 $\pm$ 6.3% -> 100.0 $\pm$ .1 | Improved |
| Add SGM to SPTM | SPTM · ViZDoom | success rate: 39.3 $\pm$ 4.0% -> 60.7 $\pm$ 4.0% | Improved |
| Add SGM to ConSPTM | ConSPTM · SafetyGym | success rate: 68.2 $\pm$ 4.1% -> 92.9 $\pm$ 1.4 | Improved |
| Add SGM and 114K cleanup steps (ablation) | SPTM · ViZDoom | overall success rate: 39.3% -> 60.7% | Improved |
| Increase cleanup from 54K to 114K steps (ablation) | SPTM + SGM · ViZDoom | overall success rate: 55.3% -> 60.7% | Improved |
| Add cleanup to sparse graphical memory (ablation) | SGM · Navigation environments | mean performance: 2.4x | Improved |
| Use two-way consistency instead of one-way consistency or uniform random subsampling (ablation) | SGM · SafetyGym | success rate | Improved |
| Use two-way consistency instead of perceptual consistency alone (ablation) | SGM · SafetyGym | success rate | Improved |
| Use sparse graphical memory for action selection and planning | SGM · PointEnv | time to take action (s): 0.550 $\pm$ 0.220 -> 0.077 $\pm$ 0.004 | Improved |
Closest papers by meaning
- Long-Horizon Visual Planning with Goal-Conditioned Hierarchical Predictors (NeurIPS 2020 · similarity 0.89)
- Sparse Planning in Visual World Models via Cost Gradients (NeurIPS 2026 · similarity 0.88)
- Visual Planning: Let's Think Only with Images (ICLR 2026 · similarity 0.88)
- DrM: Mastering Visual Reinforcement Learning through Dormant Ratio Minimization (ICLR 2024 · similarity 0.87)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Sliding Puzzles Gym: A Scalable Benchmark for State Representation in Visual Reinforcement Learning (ICML 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.
