作者位次 #4/7ACL 2026Poster· Natural Language Processing
When Is Thinking Enough? Early Exit via Sufficiency Assessment for Efficient Reasoning
Yang Xiang · Yixin Ji · Ruotao Xu · Dan Qiao · Zheming Yang · Juntao Li · Min Zhang
摘要
Large reasoning models (LRMs) have achieved remarkable performance in complex reasoning tasks, driven by their powerful inference-time scaling capability.However, LRMs often suffer from overthinking, which results in substantial computational redundancy and significantly reduces efficiency.Early-...