sifting through submitted papers (bidding) for a top ml conference & struck by how many papers are on audits/evaluation. this type of work would have remained at the fringe a few years ago. i only looked at titles ofc, it doesn’t mean any of this work can contribute to any meaningful accountability