作者位次 #10/12ACL 2026Poster· Natural Language Processing
T iny J udge: Unverifiable Constraint Alignment via Lightweight Specialist Ensembles
Yirong Zeng · Yufei Liu · Xiao Ding · Yutai Hou · Yuxian Wang · Wu Ning · Haonan Song · Dandan Tu · Qixun Zhang · Yuxiang He · Bibo Cai · Ting Liu
摘要
Instruction Following (IF) is a core capability of LLMs, requiring strict adherence to diverse constraints, ranging from verifiable ones (e.g., output length) to unverifiable ones (e.g., tone). Reinforcement learning with verifiable rewards has emerged as a paradigm for IF tasks, leveraging LLM-a...