Pat 结构实证实验包 E1-E13: 直觉路径剥离、锚绑定、succ 矩阵与差异题 (Empirical Package: Intuition-Path Stripping, Anchor Binding, Succ Matrix, Discrimination)

Empirical experiment package E1-E13 for the Pat structure: 13 reproducible experiments (scripts + results) on representation learning of true numerals (basepoint succ chains). Key findings: (1) all OOD 1.0 results are intuition paths (representation-bound statistics) - cross-representation stripping collapses them (I7n); (2) operations are anchor-bound - basepoint drift does not transfer operations, explicit teleport equivalence does not bridge across anchors, but same-anchor succ equivalence declarations DO bridge (0/32 -> 32/32, I7q); (3) anchor is the structural identity, succ is a translatable notation under the same anchor; (4) discrimination samples (2x2 vs 2+2, 0x2 vs 0+2) confirm operation discrimination extrapolates within representation (56/56); (5) token fixes: value_zero derived from basepoint, iterate_from_base serialized, sign family moved to S layer. Conjecture I7s: constructing Pat1 from two Pat0s via declaration. 中文: 13 个可复现实验: 所有 OOD 1.0 为直觉路径 (跨表示剥离全崩); 运算是锚绑定的; 同锚 succ 等价声明可桥接 (0/32→32/32); 差异题区分运算 (56/56); 猜想: 两个 Pat0 经声明产生 Pat1.