SeCuRepair: Semantics-Aligned, Curriculum-Driven, and Reasoning-Enhanced Vulnerability Repair Framework

The rapid accumulation of software vulnerabilities has outpaced manual remediation, creating an urgent need for Automated Vulnerability Re-pair (AVR). However, existing methods suffer from syntactic overfitting, mimicking surface forms without understanding the underlying repair logic, and fail to generalize to complex fixes. To transcend these limitations, we pro-pose SeCuRepair, a reliable, scalable, and efficient RL-based AVR framework. By introducing a semantic-aware reward, SeCuRepair optimizes for code semantic equivalence rather than lexical mimicry. Furthermore, SeCuRe-pair incorporates an expert-aligned reasoning mechanism that explicitly grounds patch generation in a structured diagnosis. Finally, Se-CuRepair introduces a difficulty-based curriculum that progressively disentangles the optimization barriers of entangled multi-hunk re-pairs. Extensive evaluations on a rigorous repository-level split show that SeCuRepair substantially outperforms state-of-the-art base-lines, as confirmed by both automatic evaluation and human study.