Smoothing task difficulty helps reinforcement learning models learn better

Original: Excited to share our ICML paper on RLVR learning dynamics!

Source: x.com ↗