Research map / Reinforcement learning / Offline RL
Value Memory Graph: A Graph-Structured World Model for Offline 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 |
|---|---|---|---|
| Use VMG instead of the strongest listed baseline | offline RL policy · kitchen-partial | normalized score: 49.8 -> 68.8 | Helped |
| Use VMG instead of the strongest listed baseline | offline RL policy · antmaze-large-play | normalized score: 39.6 -> 67.3 | Helped |
| Use VMG instead of the strongest listed baseline | offline RL policy · door-human | normalized score: 9.9 -> 1.5 ± 0.5 | Hurt |
| Add contrastive loss to the metric-space training (ablation) | VMG · kitchen-partial | normalized score: 0.0 -> 68.8 | Helped |
| Add contrastive loss to the metric-space training (ablation) | VMG · antmaze-medium-play | normalized score: 0.0 -> 82.7 | Helped |
| Add contrastive loss to the metric-space training (ablation) | VMG · pen-human | normalized score: 41.2 -> 70.7 | Helped |
| Add the action decoder to metric-space training (ablation) | VMG · kitchen-partial | normalized score: 15.4 -> 68.8 | Helped |
| Add multi-step search followed by Dijkstra path planning (ablation) | VMG · kitchen-partial | normalized score: 39.8 -> 68.8 | Helped |
| Add multi-step search followed by Dijkstra path planning (ablation) | VMG · pen-human | normalized score: 72.1 -> 70.7 | No clear effect |
| Replace distance-threshold state merging with K-means (ablation) | VMG · kitchen-partial | normalized score: 68.8 -> 18.3 | Hurt |
| Use the original graph reward instead of max, sum, or partial internal-reward variants (ablation) | VMG · kitchen-partial | normalized score: 31.5, 50.3, 55.0, 54.6, 67.0 -> 68.8 | Helped |
| Increase metric-space margin from 1 to 2 (ablation) | VMG · antmaze-medium-play | normalized score: 82.7 -> 84.0 | Helped |
Closest papers by meaning
- Is Value Learning Really the Main Bottleneck in Offline RL? (NeurIPS 2024 · similarity 0.89)
- The Edge-of-Reach Problem in Offline Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models (NeurIPS 2025 · similarity 0.88)
- Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining (ICLR 2025 · similarity 0.88)
- A Clean Slate for Offline Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Enhancing Value Function Estimation through First-Order State-Action Dynamics in Offline Reinforcement Learning (ICML 2024 · similarity 0.88)
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.