Mixed-difficulty reinforcement learning can create its own training curriculum

Original: 🚀 Can RLVR models find their own frontier?

Source: x.com ↗