Beyond Facts: Benchmarking Distributional Reading Comprehension in Large Language Models

Develop a benchmark tool that measures an LLM's ability to extract 'distributional knowledge'—inferring population-level trends from large corpora. Current RAG focuses on local facts; this addresses the broader analytical capacity.

Suggested repo: DistribBench

"Move beyond facts: test if your LLM actually understands trends."

Estimated effort: 40h