RAX: Scaling Online Exploration for Robot Agents

A research team from Harvard University, NVIDIA, University of California, Berkeley, Massachusetts Institute of Technology, and Stanford University has introduced RAX (Robot Agent eXploration), a framework that enables robots to learn through online exploration, achieving significant performance improvements across multiple benchmarks without requiring human demonstrations. RAX achieves an average lead of 37.4 percentage points across five simulation benchmarks and can perform tasks such as precise hole insertion and manipulation of flexible objects on real robots without human demonstration. This study marks the first systematic investigation into cross-simulation, cross-robot, and cross-embodiment online exploration scaling, providing an extensible platform for the continued development of embodied intelligence.

Researchers from Harvard, NVIDIA, UC Berkeley, MIT, and Stanford introduce RAX, a robot learning framework that enables continuous self-improvement through online exploration. RAX allows robots to acquire and reuse skills without human demonstrations, achieving significant performance gains across multiple benchmarks and real-world tasks. The system uses code-as-policy and skill accumulation, enabling robots to adapt to new environments and tasks without retraining or fine-tuning. RAX demonstrates average improvements of 37.4 percentage points over existing state-of-the-art models, with success rates reaching 98.5% on some tasks.

Researchers from Harvard, NVIDIA, UC Berkeley, MIT, and Stanford introduce RAX, a robot learning framework that enables agents to autonomously explore and accumulate skills through online exploration. RAX achieves a 37.4% average improvement over previous state-of-the-art models across five benchmarks, and demonstrates zero-shot performance on real robots without human demonstrations. The system uses code-as-policy to convert exploration into reusable skills, enabling cross-embodiment learning and deployment.

Agility Robotics announced that Digit 5 will standardize on NVIDIA IGX Thor for onboard compute, expanding its collaboration with NVIDIA as it prepares for broader commercial deployment. The transition unifies compute architecture and integrates NVIDIA Halos into safety systems, supporting industrial use and certification efforts.