Verified Events
DateTypeCompanyHeadlineSegment
AgiBot introduces WorldOracle benchmark for action-conditioned world models
- The 335-example evaluation spans AgiBot G2, YAM and human-hand embodiments with successful, failed and extended interactions.
- Seven world models are tested across rigid, deformable, fluid and articulated-object interactions.
- Metrics assess physical-state correctness, spatial alignment, interaction-area fidelity and coherence across synchronized views.
DexForce coauthors UNITAS 3D-native world action model
- The paper reports 99.8% LIBERO, 89.1% LIBERO-Plus and 92.94% / 93.04% RoboTwin Clean / Randomized success.
- Its real-robot Table 2 reports 18/20 book-organizing and 16/20 ball-packing trials, averaging 85% over these two tasks.
- The public project includes interactive scene predictions; checkpoints, inference and training code remain planned releases.