Rigorous justification of the replica-symmetric characterization

Establish a rigorous proof of the replica-symmetric ansatz and its stability for the proportional-asymptotic characterization of the L1-penalized support vector machine, or identify conditions under which the replica-symmetric solution is rigorously justified.

Background

The asymptotic results rely on a replica-symmetric calculation. The paper states that it does not provide a rigorous proof of the replica-symmetric ansatz or an Almeida–Thouless-type stability analysis, so the validity of the resulting saddlepoint equations is supported empirically rather than mathematically established.

References

Several directions remain open. First, the present theory is restricted to balanced, linearly separable data; extending it to class-imbalanced designs \citep{Loffredo2024RestoringBalance} or to the nonseparable perceptron regime \citep{Franz2019CriticalJammed} is a natural next step. Second, the theoretical framework is based on a replica-symmetric characterization and does not provide a rigorous proof of the RS ansatz or its stability; establishing such results, or identifying conditions under which the RS solution can be justified rigorously, is an important theoretical problem.

— High-Dimensional Statistical Inference for Sparse Support Vector Machines  (2610.08345 - Zeng et al., 6 Oct 2026) in Conclusion, Section 5