Kexin Chu

As large language models (LLMs) improve, many applications are moving from a single LLM call to multi-agent systems. These systems often rely on either hand-designed or automatically optimized workflows with multiple verification and testing steps. While those extra steps can improve accuracy, th...