ModelsBuildersMulti-source22:48Chinese contentReTerm:自我修复终端智能体训练方法ReTerm能从自身错误中学习,终端任务恢复能力提升显著#ReTerm#Qwen3.5-9B#Terminal-Bench#GRPO2 个信源在谈EventaarXiv cs.LG@Wei Xie et al. (6)3 sourcesSource稍后读已读值得跟进有用关注 ReTerm