This really sucks when some papers don't come with code that can reproduce the benchmark results. I wish there was a filter for "reproducable" in search results