Nvidia's Eureka helps robot dog perfect yoga ball balance

Nvidia has developed DrEureka, an LLM agent that crafts code to train robots in simulations, effectively bridging the gap between simulation and reality. This technology has enabled a robot dog to skillfully balance and walk on a yoga ball, demonstrating impressive real-world robustness even on unpredictable terrain.