Reproduce “CRONOS: Benchmarking Non-Episodic Reinforcement Learning for Robotic Manipulation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Po-Yi Wu, Djengo Cyun-Jyun Fang, Dian Cheng, Tsung-Wei Ke
The authors introduce CRONOS, a simulation benchmark for fine-tuning pretrained vision-language-action models with reinforcement learning across shared-scene tasks. It evaluates two pretrained models with PPO. In four-task sequences, policies trained with scene resets every 320 interaction steps outperform those reset every 80 on later tasks, including unseen sequences. These results are from simulation and do not establish reduced training time or interaction steps.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.