List-Decodable Coded Computing: Breaking the Adversarial Toleration Barrier

We consider the problem of coded computing, where a computational task is performed in distributed fashion presence adversarial workers. propose techniques to break toleration threshold barrier previously known computing. More specifically, we leverage list-decoding for folded Reed-Solomon codes and novel algorithms recover correct codeword using side information. In computing setting, show how master node can perform certain carefully designed extra computations obtain The workload this information negligible compared done by each worker. This then utilized prune output list decoder uniquely true outcome. further Lagrange (FLCC) incorporate developed into specific setting. Our results that FLCC outperforms LCC breaking on number adversaries be tolerated. particular, corresponding improved factor two asymptotically LCC.