arXiv · Software Engineering· Yunbo Long, Guangya Hao, Yuhan Liu, Yiting Duan, Longyan Tan, Yunchen Long, Hao Wu·2026-10-07 12:00· 3 小时前AI 评分35

编程智能体的自纠正应承载多少证据？面向自蒸馏的自适应 Dirichlet 证据

How Much Evidence Should a Coding Agent's Self-Correction Carry? Adaptive Dirichlet Evidence for Self-Distillation

AI 导读

论文提出 Effective-Evidence Self-Distillation（EESD），把执行反馈中的转移支持与证据质量分开建模，用 Dirichlet 后验生成带不确定性惩罚的权重进行 KL 锚定的纠正学习。