> ...if we require every paper to beat state-of-the-art, we'd (collectively, as the entire discipline) be lucky to publish one paper a year. That's the case if

> ...if we require every paper to beat state-of-the-art, we'd (collectively, as the entire discipline) be lucky to publish one paper a year. That's the case if there's just one baseline to beat. It would be a bad thing if everyone were working on beating today's single state of the art model. It would be great to incentivize more work along varied lines. Either way, we absolutely don't want papers that claim their innovation is an improvement on a metric they don't really improve.