Welcome to the AI research bites. This series of short and informative talks showcases cutting-edge research work from ServiceNow AI Research team. The AI Research Bites are open to all, especially those interested in keeping up with the fast-paced AI research community.  PipelineRL accelerates Reinforcement Learning by 2x through parallelizing data generation and training, eliminating GPU idle time via continuous batching and in-flight weight updates. In this presentation, Rafael Pardinas demonstrates that decoupling inference from optimization enables constant batch sizes and significantly higher throughput than conventional RL pipelines. The approach maintains strong sample efficiency and matches the learning dynamics of recent SOTA RL methods, demonstrating that training speed and on-policy effectiveness do not need to be mutually exclusive.  Paper: https://arxiv.org/pdf/2509.19128 ServiceNow AI Research team: https://www.servicenow.com/research/

 297      10