The Distillation-to-Reinforcement Learning (D2RL) method is proposed to address the limited generalization of supervised fine-tuning and the sparse reward problem in reinforcement learning, and produces substantial gains over models trained on open-source datasets, with performance comparable or even superior to models...
Qing-Song Xie, Jia-Hao Tian, Jia-Cheng Tang et al.· 0 citations
AutoResearchEval, featuring 100 tasks grounded in published frontier science across 7 scientific domains and the full research lifecycle, and ARFT, a framework of 45 empirically-grounded failure patterns to enable scalable fine-grained attribution, are introduced.
This work shows that a state-of-the-art weight-inheritance distiller, Low-Rank Clone (LRC), deploys a full-width student MLP but ties training to a teacher-induced slice, leaving 62.5-81.4% of each deployed matrix's independent linear degrees of freedom unreachable-paid for at inference, never trainable.
Wen-Hui Chen, Zhi-Feng Li, Jie Zhou et al.· 0 citations
Reconstruction is introduced, a blind idea-recovery benchmark that withholds the seed paper and all contemporaneous or future literature, and asks models to propose hypotheses that an independent large language model judge matches against the held-out ground-truth idea.
Shao-Long Chen, Yanlin Fei, Nazhou Liu et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
