## Problem 143 — rank:88

**Title:** Higher-order epistasis in evolution

**Added binding assumptions (LaTeX):**

```latex
The Walsh expansion is with respect to a declared measure on the full genotype cube, even when measurements cover only part of it. A small spectral tail gives an $L^2$ approximation under that measure, not automatically preservation of every selection-gradient sign or accessible path. Such claims need a pointwise error bound and a positive fitness-difference margin, or an explicit probability of path misclassification under the chosen evolutionary law.
```

**Direct edits:** 0. **Fresh external checks:** None; supplied sources retained as provenance.

