一作ACL 2026Poster· Natural Language Processing
Reliable Use of Lemmas via Eligibility Reasoning and Section-Aware Reinforcement Learning
Zhikun Xu · Xiaodong Yu · Ben Zhou · Jiang Liu · Jialian Wu · Ze Wang · Ximeng Sun · Hao Chen · Zicheng Liu
摘要
Recent large language models (LLMs) perform strongly on mathematical benchmarks yet often misapply lemmas, importing conclusions without validating assumptions. We formalize lemma-judging as a structured prediction task: given a statement and a candidate lemma, the model must output a preconditio...