1
Extrapolative Weight Averaging Reveals Correctness-Efficiency Frontiers in Code RL
外推权重平均法揭示了代码强化学习中正确性与效率的权衡前沿，为模型调优提供新视角。
arXiv:2605.28751v1 Announce Type: cross Abstract: Linear interpolation between fine-tuned checkpoints has been shown to trace the Pareto front between…