Research Method Comparison Agent

Literature comparisons of research methods often rank them by headline numbers that came from different datasets, metrics or setups, which is misleading. This agent extracts each paper's stated conditions: dataset, metric, test split, hardware and key settings. Where conditions differ in ways that might explain a performance gap, it reads the methods and appendix more closely. It then builds a methods matrix that shows results with their conditions. Methods evaluated on substantially different datasets are never ranked against each other; the matrix says plainly that the comparison is not direct. The research lead reviews any conclusions drawn from the matrix. Edge case: two papers that both report accuracy but define it differently are treated as not comparable.