Research map / Video / Video generation
Walk through Paintings : Ego-centric World models from Internet Priors
Anurag Bagchi, Zhipeng Bao, Homanga Bharadhwaj, Yu-Xiong Wang, Pavel Tokmakov, Martial Hebert
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 pretrained Internet video weights (vs training from scratch) (ablation) | EgoWM (SVD) · 1X humanoid navigation, frame 16 | SCS: 21.6 -> 34.4 | Improved |
| add pretrained Internet video weights (vs training from scratch) (ablation) | EgoWM (SVD) · 1X humanoid navigation, frame 16 | LPIPS: 0.47 -> 0.35 | Improved |
| use temporally aligned action injection instead of global conditioning (ablation) | EgoWM (Cosmos-2B) · 1X humanoid navigation (16 frames, k=4) | SCS: 14.9 -> 27.0 | Improved |
| use temporally aligned action injection instead of chunked conditioning (ablation) | EgoWM (Cosmos-2B) · 1X humanoid navigation (16 frames, k=4) | SCS: 23.7 -> 27.0 | Improved |
| use temporally aligned action injection instead of chunked conditioning (ablation) | EgoWM (Cosmos-2B) · 1X humanoid navigation (16 frames, k=4) | DreamSim: 0.17 -> 0.19 | No clear effect |
| use lightweight action conditioning instead of cross-attention (ablation) | EgoWM (SVD) · RECON OOD (16 frames) | SCS: 22.7 -> 29.7 | Improved |
| use lightweight action conditioning instead of cross-attention (ablation) | EgoWM (SVD) · RECON OOD (16 frames) | LPIPS: 0.64 -> 0.59 | Improved |
| replace NWM with EgoWM (SVD) | NWM · RECON val 3-DoF navigation, frame 16 | SCS: 33.4 -> 55.2 | Improved |
| replace NWM with EgoWM (Cosmos-2B, k=4) | NWM · RECON val 3-DoF navigation, frame 16 | LPIPS: 0.45 -> 0.33 | Improved |
| use temporal compression k=4 (Cosmos/Wan) instead of SVD k=1 | EgoWM · RECON val 3-DoF navigation, frame 16 | SCS: 55.2 -> 47.5 | Mixed |
| replace NWM with EgoWM (Cosmos) for planning | NWM · RECON 3-DoF planning | ATE: 3.25 -> 3.13 | Improved |
| replace NWM with EgoWM (SVD) in 64-frame rollouts | NWM · 3-DoF navigation 64-frame rollouts | SCS: 0.15 -> 0.41 | Improved |
Closest papers by meaning
- Navigation World Models (CVPR 2025 · similarity 0.90)
- WorldSimBench: Towards Video Generation Models as World Simulators (ICML 2025 · similarity 0.90)
- RoboScape: Physics-informed Embodied World Model (NeurIPS 2025 · similarity 0.89)
- Learning 3D Persistent Embodied World Models (NeurIPS 2025 · similarity 0.89)
- Grounding Video Models to Actions through Goal Conditioned Exploration (ICLR 2025 · similarity 0.89)
- DriveDreamer-Policy: A Geometry-Grounded World-Action Model for Unified Generation and Planning (NeurIPS 2026 · similarity 0.89)
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.
