SoRFT: Issue Resolving with Subtask-oriented Reinforced Fine-Tuning
arXiv:2502.20127v1 Announce Type: cross Abstract: Mainstream issue-resolving frameworks predominantly rely on commercial models, leading to high costs and privacy concerns. […]
SoRFT: Issue Resolving with Subtask-oriented Reinforced Fine-Tuning Lire l’article »