Doing simple retrieval from LLM models at various context lengths to measure accuracy - View it on GitHub