Odyssey transfers a GTA-trained policy to RDR2 horseback riding without retraining
nathanbenaich · x · 2026-10-15
Nathan Benaich (State of AI) highlights Odyssey-ML's thesis: learning to predict the world may also help models act in it, explored across driving, robotics, and games. One example: a policy trained in GTA transferred to horseback movement in Red Dead Redemption 2 with zero retraining — evidence for world-prediction as a basis for general policy transfer.
More from Embodied
- WEDGE uses navigation agents' own failures to generate targeted training environments — EliasEskin · 2026-10-16
- Sandbar's Stream Ring named a TIME Best Invention of 2026 in wearables — km · 2026-10-15
- Nvidia Presents Robota-Embed: Unified VLA Embeddings for Robot In-Context Learning — arankomatsuzaki · 2026-10-15
- Chinese startup DexForce trains humanoid robots with virtual data to slash costs — pstAsiatech · 2026-10-15
- EmbodiedEvo tops RoboDojo at 38.95% success by letting robots inherit failed practice — thetripathi58 · 2026-10-15
- VR dev's one line stuns a Quest 3 owner who played 3 games for 12 months ignoring mixed reality — Aiden_Tech_Ai · 2026-10-15