← back
arXivJian Mu, Tianyi Lin, Chengwei Qin, Zhongxiang Dai, Yao ShuFri, May 29, 2026, 8:49 AM PDT
score 14.7

Efficient multi-turn AI training without expensive real-time learning

Original: DRIFT: Decoupled Rollouts and Importance-Weighted Fine-Tuning for Efficient Multi-Turn Optimization

Source: arxiv.org

Writing ELI5 summary…