Relaxed Local Correctability from Local Testing

View PDF

We cement the intuitive connection between relaxed local correctability and local testing by presenting a concrete framework for building a relaxed locally correctable code from any family of linear locally testable codes with sufficiently high rate. When instantiated using the locally testable codes of Dinur et al. (STOC 2022), this framework yields the first asymptotically good relaxed locally correctable and decodable codes with polylogarithmic query complexity, which finally closes the superpolynomial gap between query lower and upper bounds. Our construction combines high-rate locally testable codes of various sizes to produce a code that is locally testable at every scale: we can gradually "zoom in" to any desired codeword index, and a local tester at each step certifies that the next, smaller restriction of the input has low error.
Our codes asymptotically inherit the rate and distance of any locally testable code used in the final step of the construction. Therefore, our technique also yields nonexplicit relaxed locally correctable codes with polylogarithmic query complexity that have rate and distance approaching the Gilbert-Varshamov bound.