Research map / Reinforcement learning / Visual and representation RL
Learning Dynamic Attribute-factored World Models for Efficient Multi-object 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 attribute-factored dynamics and rewards with class templates, interaction patterns, dynamic interaction graphs, and latent-parameter inference rather than the strongest listed baseline | DAFT-RL (Symbolic) · Push & Switch compositional generalization, excluding 2-Push + 2-Switch (S) | success rate: 0.751 -> 0.805; 0.964 -> 0.968; 0.919 -> 0.923; 0.915 -> 0.921; 0.634 -> 0.793; 0.531 -> 0.783 | Helped |
| Use DAFT-RL instead of NCS | DAFT-RL (Symbolic) · 2-Push + 2-Switch (S) | success rate: 0.912 -> 0.881 | Hurt |
| Use DAFT-RL with Slot Attention instead of the strongest listed baseline | DAFT-RL (SA) · Spriteworld object goal, interaction, comparison, and property comparison | success rate: 0.847 -> 0.897; 0.837 -> 0.890; 0.834 -> 0.893; 0.831 -> 0.907 | Helped |
| Use DAFT-RL with AIR instead of the strongest listed baseline | DAFT-RL (AIR) · Block-stacking, 2 and 4 blocks | success rate: 0.797 -> 0.838; 0.605 -> 0.698 | Helped |
| Use DAFT-RL with AIR instead of LRN | DAFT-RL (AIR) · Block-stacking, 6 and 8 blocks | success rate: 0.602 -> 0.664; 0.463 -> 0.571 | Helped |
| Use DAFT-RL with Slot Attention instead of NCS | DAFT-RL (SA) · Block-stacking, 2 blocks | success rate: 0.797 -> 0.809 | Helped |
| Remove latent-parameter modeling (ablation) | DAFT-RL · 2-Push + 2-Switch (L+S) | success rate | Hurt |
| Remove class template graphs (ablation) | DAFT-RL · 2-Push + 2-Switch (L+S) | success rate | Hurt |
| Remove the dynamic interaction graph (ablation) | DAFT-RL · 2-Push + 2-Switch (L+S) | success rate | Hurt |
| Remove interaction pattern graphs (ablation) | DAFT-RL · 2-Push + 2-Switch (L+S) | success rate | Hurt |
Closest papers by meaning
- Efficient Exploration and Discriminative World Model Learning with an Object-Centric Abstraction (ICLR 2025 · similarity 0.89)
- Entity-Centric Reinforcement Learning for Object Manipulation from Pixels (ICLR 2024 · similarity 0.88)
- Learning Interactive World Model for Object-Centric Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Modeling Unseen Environments with Language-guided Composable Causal Components in Reinforcement Learning (ICLR 2025 · similarity 0.88)
- SOLD: Slot Object-Centric Latent Dynamics Models for Relational Manipulation Learning from Pixels (ICML 2025 · similarity 0.87)
- On Generalization Across Environments In Multi-Objective Reinforcement Learning (ICLR 2025 · 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.