Abstract page for arXiv paper 2610.12545: AutoIndexer: Flexible-Order Decoding by Training Causal Models on Chains of Edits