KCIF is a benchmark for evaluating the instruction-following capabilities of Large Language Models (LLM). - View it on GitHub