作者位次 #3/7ACL 2026Poster· Natural Language Processing
PARIF : Pushing the P areto Frontier of Instruction Following and Reasoning with Curriculum Reinforcement Learning
Rongchuan Mu · Zexin Wang · Qianyu Wang · MingHua Ma · Zekun Wang · Ming Liu · Bing Qin
摘要
Large Reasoning Models (LRMs) excel at complex problem-solving but frequently overlook specific instruction constraints. Existing alignment methods struggle to balance general reasoning with instruction-following (IF), hindered by dependency on teacher models, reward hacking, and reasoning-answer...