Introduces ARISE, a hierarchical reinforcement learning framework that allows LLMs to evolve and reuse a tiered library of reasoning skills rather than treating every math problem in isolation.
arXiv · March 18, 2026 · 2603.16060
The Takeaway
It shifts LLM reasoning from simple token prediction to 'skill management,' allowing models to summarize successful solution traces into reusable strategies that improve OOD performance.
From the abstract
The dominant paradigm for improving mathematical reasoning in language models relies on Reinforcement Learning with verifiable rewards. Yet existing methods treat each problem instance in isolation without leveraging the reusable strategies that emerge and accumulate during training. To this end, we introduce ARISE (Agent Reasoning via Intrinsic Skill Evolution), a hierarchical reinforcement learning framework, in which a shared policy operates both to manage skills at high-level and to generate