Convex Compositional Reasoning Models
Abstract
Compositional energy-based models can generalize to larger combinatorial reasoning problems by reusing a learned factor energy across many local constraints. In our paper, we show that a key bottleneck in compositional reasoning is not composition itself, but the non-convex geometry of the learned energy landscape. To solve this problem, we introduce Convex Compositional Energy Minimization (CCEM), a framework that parameterizes each factor with an input-convex neural network and optimizes the composed energy over a tight convex relaxation of the feasible set. Because convexity is preserved under summation, the global relaxed objective remains convex, enabling deterministic projected first-order optimization. CCEM is trained in two stages: factor-level contrastive learning to shape local energy basins, followed by end-to-end refinement through an unrolled projected solver. Our experiments show that our models trained on small subproblems or a single problem size transfer to larger instances without retraining.
Cite
@article{arxiv.2605.23395,
title = {Convex Compositional Reasoning Models},
author = {Meir Roketlishvili and Semyon Semenov and Maksim Bobrin and Viktor Kovalchuk and Albert Baichorov and Abduragim Shtanchaev and Fakhri Karray and Dmitry V. Dylov and Martin Takáč and Arip Asadulaev},
journal= {arXiv preprint arXiv:2605.23395},
year = {2026}
}