@article{trajectorybalancewithasynchronydecouplin, title = {Trajectory Balance with Asynchrony: Decoupling Exploration and Learning for Fast, Scalable LLM Post-Training}, author = {Brian R. Bartoldson and Siddarth Venkatraman and James Diffenderfer and Moksh Jain and Tal Ben-Nun and Seanie Lee and Minsu Kim and Johan Obando-Ceron and Yoshua Bengio and Bhavya Kailkhura}, year = {2025}, eprint = {2503.18929}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2503.18929v1}, }