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

Open in new window

The document is on a web server that cannot be reached. Please try again later.