一作ICLR 2026Poster· Deep Learning / Everything Else
The Pensieve Paradigm: Stateful Language Models Mastering Their Own Context
Xiaoyuan Liu · Tian Liang · Dongyang Ma · Deyu Zhou · Haitao Mi · Pinjia He · Yan Wang
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Xiaoyuan Liu · Tian Liang · Dongyang Ma · Deyu Zhou · Haitao Mi · Pinjia He · Yan Wang
Youliang Yuan · Qiuyang Mang · Jingbang Chen · Hong Wan · Xiaoyuan Liu · Junjielong Xu · Jen-Tse Huang · Wenxuan Wang · Wenxiang Jiao · Pinjia He
In this paper, we observe that current models are susceptible to reward hacking, leading to a substantial overestimation of a model’s reasoning ability. This is evidenced by a high incidence of “false positives”—solutions that reach the correct answer through an unsound process.Through a systemat...