作者位次 #4/7ACL 2026Poster· Natural Language Processing
Powering Verifiable Learning via Automated Evolutionary Data Synthesis
He Du · Bowen Li · Aijun Yang · Siyang He · Qipeng Guo · Kai Chen · Dacheng Tao
摘要
Reliable verifiable data has become a key driver of capability gains in modern language models, enabling stable reinforcement learning with verifiable rewards and effective distillation that transfers competence across math, coding, and agentic tasks. Yet constructing generalizable synthetic veri...