TarrasqueSNT: Adaptive Pre-Decoding Error Shaping for Quantum Fault Tolerance

We introduce TarrasqueSNT, an adaptive pre-decoding transformation framework for quantum error correction that reshapes error distributions into configurations more favorable for minimum-weight perfect matching (MWPM) decoding, without modifying the decoder itself. Three structured operators are mapped onto quantum error correction primitives: an iterative Cyclic Reset operator with adaptive probability schedule γc (p) ; a Diversifier that performs L0 → L1 subspace transformation of uncorrectable syndromes; and a Liminal gate that routes syndrome measurements based on reliability. Simulations on rotated surface codes (d = 3, d = 5) using Stim and PyMatching confirm consistent logical error rate reductions of 20–35% over standard MWPM across p = 0. 005–0. 020 (N = 100, 000 shots). At practical noise regimes (p ≥ 0. 007), TarrasqueSNT d = 3 achieves lower logical error rates than standard MWPM d = 5, suggesting an effective reduction in distance requirements with 64% fewer physical qubits. An ablation study reveals that the Cyclic Reset operator accounts for ∼73% of total gains, while runtime measurements confirm no measurable computational overhead — TarrasqueSNT is faster than standard MWPM. These results establish pre-decoding error shaping as a new optimization axis in quantum fault tolerance, complementary to advances in code design and decoder algorithms.