AllenAI released ToyBlocks, an open simulation framework and dataset for robot learning research using a robot arm and tabletop blocks. The release includes 85,336 long-horizon demonstrations, about 1,600 hours of robot execution data, and 144 million frames, along with simulation assets, task and demonstration generators, two trained models, training code, and evaluation code and validation sets.
ToyBlocks supports task generation, model training, and evaluation with measures such as execution speed, forces, and task progress. AllenAI said a larger dataset, more skills, broader benchmarking, and integrated reinforcement learning training are planned for a later release; integrated RL training is not part of the current package.
AI 根据报道生成 · 2 小时前更新
