作者位次 #2/3ACL 2026Findings· Findings of ACL
A bstain- R 1: Calibrated Abstention and Post-Refusal Clarification via Verifiable RL
Haotian Zhai · Jingcheng Liang · Dongyeop Kang
摘要
Reinforcement fine-tuning improves the reasoning ability of large language models, but it can also encourage them to answer unanswerable queries by guessing or hallucinating missing information. Existing abstention methods either train models to produce generic refusals or encourage follow-up cla...