AI ByteDance AI Research Unveils Reinforced Fine-Tuning (ReFT) Method to Enhance the Generalizability of Learning LLMs for Reasoning with Math Problem Solving as an Example
