The question
How much apparent code advantage remains under optimal decoding, and when can edge reweighting recover performance under correlated noise?
Why it matters
An error-correcting code and its decoder are different parts of a reliability system. A decoder that assumes independent errors can miss structure in real noise, making a code look worse than it is.
What you could investigate
- Compare exact maximum-likelihood decoding with a graph-based decoder such as minimum-weight perfect matching on small surface-code instances.
- Introduce biased noise and tunable nearest-neighbor correlations, including paired Z errors.
- Measure the decoding-performance gap and test whether effective edge reweighting can close it.
What you could create
- A controlled comparison of code and decoder effects
- Performance maps under bias and correlation
- An account of where reweighting succeeds or fails