Fetching the paper…

RL on Incorrect Synthetic Data Scales the Efficiency of LLM Math Reasoning by Eight-Fold · Around