ReSI-Math: Exploring Recoverable Capability Frontiers for Efficient Reasoning Improvement
Reinforcement learning with verifiable rewards (RLVR) has become an effective approach for improving mathematical reasoning in large language models. However, a fundamental question remains: when a model fails to solve a problem, does it truly lack the underlying capability? In this work, we reveal that direct failure...
Yao Cui, Zhen Yang, Yuan-Dong Wang et al.
· 0 citations