Benchmark of stylistic variation in LLM-generated texts