Towards Verifiable Generation: A Benchmark for Knowledge-aware Language Model Attribution