Here we provide models and benchmarks for the Just Read Twice work: https://arxiv.org/abs/2407.05483