Research map / Video / Video segmentation and tracking
Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling
Tal Daniel, Carl Qi, Dan Haramati, Amir Zadeh, Chuan Li, Aviv Tamar, Deepak Pathak, David Held
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 |
|---|---|---|---|
| use object-centric particles instead of patch-based DVAE | world model · Sketchy-U | LPIPS: 0.113 -> 0.070 | Improved |
| use object-centric particles instead of patch-based DVAE | world model · Mario-U | FVD: 277.41 -> 195.95 | Improved |
| use object-centric particles instead of patch-based DVAE | world model · Bridge-L | FVD: 146.85 -> 47.78 | Improved |
| use object-centric particles instead of patch-based DVAE | world model · LanguageTable-L | FVD: 26.78 -> 15.96 | Improved |
| use object-centric particles instead of patch-based DVAE | world model · BAIR-U | FVD: 164.41 -> 163.91 | No clear effect |
| use object-centric particles instead of patch-based DVAE | world model · LanguageTable-A | LPIPS: 0.038 -> 0.037 | No clear effect |
| use per-particle latent actions instead of PlaySlot global latent action | video prediction · Sketchy-U | LPIPS: 0.275 -> 0.070 | Improved |
| use per-particle latent actions rather than global latent actions (ablation) | LPWM · video prediction | performance | Improved |
| embed positions via AdaLN instead of additive positional embeddings (ablation) | LPWM · video prediction | performance | Improved |
| vary latent action dimension near the particle dimension (ablation) | LPWM · video prediction | performance | No clear effect |
| use LPWM instead of EC Diffuser on PandaPush | imitation learning · PandaPush 2 Cubes | success rate (%): 91.7 -> 74 | Hurt |
| use LPWM instead of HIQL on OGBench-Scene | imitation learning · OGBench-Scene task3 | success rate (%): 61 -> 89 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PlaySlot — PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning (ICML 2025) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning (ICML 2025 · similarity 0.88)
- Learning Visual Feature-Based World Models via Residual Latent Action (NeurIPS 2026 · similarity 0.88)
- Zero-shot World Models via Search in Memory (NeurIPS 2025 · similarity 0.86)
- Learning 3D Particle-based Simulators from RGB-D Videos (ICLR 2024 · similarity 0.86)
- LaMo: Self-Supervised Latent Motion Priors for Physical Realism in Video Generation (NeurIPS 2026 · similarity 0.86)
- Simple, Good, Fast: Self-Supervised World Models Free of Baggage (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.
