aiqbench analysis
This research introduces Multi-Attractor Graph Neural Networks capable of representing set-valued maps by converging to multiple valid solutions rather than enforcing a single fixed point. By leveraging different initializations, these models can explore diverse solution branches for combinatorial and scientific problems without relying on arbitrary target selection or auxiliary identifiers. The approach demonstrates superior average solution quality compared to standard baselines while maintaining numerical convergence rates competitive with larger diffusion-based solvers.
Multi-Attractor GNNs: Set-Valued Expressivity Beyond Unique Equilibria
arXiv:2609.35274v1 Announce Type: new Abstract: Recurrent and equilibrium graph neural networks (GNNs) often enforce a unique fixed point or use one training target per graph. Yet many combinatorial and scientific problems admit multiple valid solutions, with no preferred one. A designated target can then impose an arbitrary selection rule. For tasks invariant to node relabeling, a symmetric graph may have a symmetric solution set but no symmetric solution. We show that multiple equilibria enable one weight-tied message-passing GNN to represent set-valued equivariant maps: different initializations approach different valid solutions. Under stated regularity assumptions, we first construct globally Lipschitz, permutation-equivariant dynamics that converge almost surely to valid solutions and reach every solution branch with positive probability. We then establish approximate realization by recurrent message passing with continuous component maps, with arbitrarily small update and limiting errors and arbitrarily high probability. This goes beyond standard universality arguments: although message passing alone cannot distinguish symmetric nodes, the evolving state keeps nodes distinguishable at every finite step without auxiliary node identifiers. Such dynamics can be learned without solution labels using problem-specific energies. On Ising ground states, structural module detection in protein graphs, and chemical reaction steady states, the learned updates produce multiple high-quality predictions with high numerical convergence rates. They achieve better average solution quality than the tested unique-equilibrium, single-target, and feedforward baselines, while remaining competitive with much larger diffusion-based solvers.
Abstract:Recurrent and equilibrium graph neural networks (GNNs) often enforce a unique fixed point or use one training target per graph. Yet many combinatorial and scientific problems admit multiple valid solutions, with no preferred one. A designated target can then impose an arbitrary selection rule. For tasks invariant to node relabeling, a symmetric graph may have a symmetric solution set but no symmetric solution. We show that multiple equilibria enable one weight-tied message-passing GNN to represent set-valued equivariant maps: different initializations approach different valid solutions. Under stated regularity assumptions, we first construct globally Lipschitz, permutation-equivariant dynamics that converge almost surely to valid solutions and reach every solution branch with positive probability. We then establish approximate realization by recurrent message passing with continuous component maps, with arbitrarily small update and limiting errors and arbitrarily high probability. This goes beyond standard universality arguments: although message passing alone cannot distinguish symmetric nodes, the evolving state keeps nodes distinguishable at every finite step without auxiliary node identifiers. Such dynamics can be learned without solution labels using problem-specific energies. On Ising ground states, structural module detection in protein graphs, and chemical reaction steady states, the learned updates produce multiple high-quality predictions with high numerical convergence rates. They achieve better average solution quality than the tested unique-equilibrium, single-target, and feedforward baselines, while remaining competitive with much larger diffusion-based solvers.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2609.35274 [cs.LG]
(or arXiv:2609.35274v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.35274 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 319 words
Related coverage
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · Oct 13, 2026
LiveLaw Supreme Court Half-Yearly Digest: January - June 2026 [Part 2]
livelaw.in · Sep 13, 2026
[AINews] OpenAI shuts off Cursor
Latent Space · Aug 29, 2026
Academia is for Ambition — Alex Zhang, MIT
Latent Space · Oct 2, 2026