Training AI to reason better through long documents

Original: LongTraceRL: Learning Long-Context Reasoning from Search Agent Trajectories with Rubric Rewards

Source: arxiv.org ↗