Apple researchers built an AI that tests several ideas in parallel before answering - 9to5Mac
… On math benchmarks, LaDiR achieved higher accuracy than existing approaches and demonstrated stronger performance even on more difficult, out-of-distribution tasks. …