Section
Physical AI Papers
- Yann LeCun's H-JEPA: 73% vs 18% Without Generating a Single Pixel (2610.06805)
H-JEPA (arXiv 2610.06805) reports 73% vs 18% success on Visual AntMaze with a three-level latent world model. What it shows, and where it fails.
- NVIDIA Long-WAM: Robot Success Jumps From 63% to 79% With 19 Seconds of Memory (2610.10528)
Long-WAM (arXiv 2610.10528) reports RoboCasa GR-1 success rising from 63.3% to 78.7% with 19.2 s of visual history. What holds up, and what doesn't.