Mosaic: multi-session agentic reinforcement learning
A paper submitted to arXiv on 15 Oct 2026, "Mosaic: Multi-Session Agentic Reinforcement Learning," introduces multi-session agentic RL, a more general setting in which a trajectory may span an arbitrary number of sessions to match inference-time execution such as context compaction and subagent forking. The authors note that the common practice of broadcasting trajectory-level advantages to all sessions neither distinguishes session contributions nor guides session allocation, so they propose Mosaic, a critic-free hierarchical method that folds session-usage signals into trajectory-level advantage estimation and shapes each session's advantage by its estimated marginal contribution. Experiments on BrowseComp with Qwen3.5-4B show Mosaic outperforming broadcasting in both context compaction and subagent forking scenarios while using fewer sessions on average.
It aligns training assumptions with how multi-session agents actually execute at inference, offering a concrete optimization target for future work.