一作ICML 2026Poster· Deep Learning / Large Language Models
DRIFT: Decoupled Rollouts and Importance-Weighted Fine-Tuning for Efficient Multi-Turn Optimization
Jian Mu · Tianyi Lin · Chengwei Qin · Zhongxiang Dai · Yao Shu
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Jian Mu · Tianyi Lin · Chengwei Qin · Zhongxiang Dai · Yao Shu
Jian Mu · Qixin ZHANG · Zhiyong Wang · Menglin Yang · Shuang Qiu · Chengwei Qin · Zhongxiang Dai · Yao Shu
Large language models (LLMs) increasingly solve complex reasoning tasks via long chain-of-thought, but their forward-only autoregressive generation process is fragile; early token errors can cascade, which creates a clear need for self-reflection mechanisms. However, existing self-reflection eith...