This approach would remove one of the main benefits - the ability to run multi-task one-shot prompts where a single LLM call returns answers to multiple NLP tasks.