Research map / Reinforcement learning / Visual and representation RL
Model-Based Visual Planning with Self-Supervised Functional Distances
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 planning with the predictive model to Q-function-based control (ablation) | MBOLD · 1 object push | success rate: 19.2 ± 3.6 % -> 55.2 ± 4.3% | Helped |
| Add planning with the predictive model to Q-function-based control (ablation) | MBOLD · 3 object push | success rate: 15.6 ± 3.6 % -> 44.8 ± 2.9% | Helped |
| Add planning with the predictive model to Q-function-based control (ablation) | MBOLD · Reach | success rate: 31.8 ± 5.2 % -> 94.4 ± 3.3% | Helped |
| Add the conservative Q-learning loss (ablation) | MBOLD · Door sliding | mean success rate: 41% -> 58% | Helped |
| Add the conservative Q-learning loss (ablation) | MBOLD · Other evaluated tasks | success rate | No clear effect |
| Restrict the distribution of object positions at episode reset (ablation) | MBOLD · 1 object push (regular and hard) | success rate: 55.2 ± 4.3% -> 54.5 ± 3.9 %; 40.2 ± 7.2% -> 43.2 ± 7.2 % | No clear effect |
| Remove negative goal mining (ablation) | MBOLD · Object pushing, object movement goals | success rate (arm position match): 64.4 ± 5.9% -> 83.4 ± 4.0 % | Helped |
| Remove negative goal mining (ablation) | MBOLD · Object pushing, no object movement goals | success rate (arm position match): 89.2 ± 1.9% -> 91.6 ± 2.3 % | No clear effect |
| Use negative goal mining (ablation) | MBOLD · Object pushing, object relocation | success rate | Helped |
| Use longer planning horizons (ablation) | MBOLD · Simulated block pushing | control performance | Helped |
| Use a reward of 1 per non-goal step and 10 upon reaching the goal instead of (0, 1) (ablation) | MBOLD distance learning · Evaluation environments | performance | No clear effect |
Closest papers by meaning
- Stabilizing Contrastive RL: Techniques for Robotic Goal Reaching from Offline Data (ICLR 2024 · similarity 0.89)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.89)
- Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models (NeurIPS 2025 · similarity 0.89)
- OSVI-WM: One-Shot Visual Imitation for Unseen Tasks using World-Model-Guided Trajectory Generation (NeurIPS 2025 · similarity 0.89)
- Efficient Robotic Policy Learning via Latent Space Backward Planning (ICML 2025 · similarity 0.88)
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · 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.