AI system trains multiple language models against each other to improve reasoning

Original: PopuLoRA: Co-Evolving LLM Populations for Reasoning Self- Play

Source: vmax.ai ↗