作者位次 #2/6ACL 2026Findings· Findings of ACL
Efficient Test-Time Scaling via Temporal Reasoning Aggregation
Jiakun Li · Xingwei He · Kefan Li · Hongzheng Chai · Hongyue Yu · Yuan Yuan
摘要
Test-time scaling improves the reasoning performance of large language models but often results in token-inefficient overthinking, where models continue reasoning beyond what is necessary for a correct answer. Existing dynamic early-exit methods typically rely on single-step confidence signals, w...