Left-Right Cayley LTC Comes From Local Agreement Plus Expansion¶
Claim/Theorem¶
Let
be a left-right Cayley complex such that both \(\operatorname{Cay}(G,A)\) and \(\operatorname{Cay}(G,B)\) are \(\lambda\)-expanders and the noncollision condition of the paper holds. Let \(C_A\subseteq\{0,1\}^A\) and \(C_B\subseteq\{0,1\}^B\) be base codes of relative distances \(\delta_A,\delta_B>0\), and suppose the tensor code \(C_A\otimes C_B\) is \(\kappa_0\)-agreement testable. Define
If
then the global square code
is \(\kappa\)-locally testable with \(|A||B|\) queries, where
Equivalently, for every word \(f\) on squares,
The proof is local-to-global:
- each vertex test checks membership in the local tensor code \(C_A\otimes C_B\);
- an iterative correction algorithm reduces local disagreement between neighboring views;
- expansion of both one-skeleton directions, together with the parallel-walk consistency argument, forces either global agreement with a nearby codeword or a constant fraction of violated local tests.
So the nearby classical code family in the left-right Cayley framework is not just an LTC in the abstract sense. It is an LTC produced by the stronger package: local tensor agreement testability plus two-direction expansion plus a local-to-global consistency mechanism.
Dependencies¶
- None.
Conflicts/Gaps¶
- The conclusion is still only global local testability. By [[good-ltc-does-not-imply-balanced-cut-rank.md]], that alone cannot force balanced-cut rank-connectivity.
- The node therefore isolates a candidate extra ingredient for the frontier, but it does not yet convert that ingredient into trellis-width, branchwidth, or intrinsic cut-rank lower bounds.
- Translating this left-right-Cayley local-to-global mechanism into the stabilizer-space quantity [[stabilizer-cut-rank-functional.md]] remains fully open.
Sources¶
10.1145/3519935.3520024