I think the README should include a couple of examples for - replicating the benchmarks from the paper - using COPRA on a custom goal (being able to customize goal, previous context in the file, model used for solving)