CodeBLEU

anchor

Findings

• IC-073Released LLMs (GPT-4o, Llama-3.1-70B, Qwen2-7B, etc.) show limited workflow orchestration capability that degrades as workflow complexity increases [eval]