Learning to Trace Seiberg Dualities
2026-07-30 • Artificial Intelligence
Artificial IntelligenceMachine Learning
AI summaryⓘ
The authors explore how machine learning can help identify when two complex physics systems, called supersymmetric quiver gauge theories, are essentially the same under a concept known as Seiberg duality. They treat this problem like learning to untangle knots in networks, using neural networks such as transformers and perceptrons. Their results show that these AI methods can often find these dualities more efficiently than traditional algorithms, especially for medium-sized networks. Adding specialized pathfinding techniques further improves the process. The authors suggest this approach could be a useful test for advanced AI in theoretical physics.
Seiberg dualitysupersymmetric quiver gauge theoryquiver mutationmachine learningtransformersmulti-layer perceptronsnetwork architecturepathfinder algorithmscomputational complexity
Authors
Jonathan J. Heckman, Shani Meynet, Alessandro Mininno, Gary Shiu
Abstract
Dualities play an important role in establishing both microscopic and emergent phenomena in a wide range of physical systems. In practice, though, it can often be computationally challenging to establish when two systems are dual, even when all of the "rules of the game" are well-known. Said differently, when confronted with two systems, how can one efficiently establish that they are in fact dual? In this paper we use machine learning methods to address this question for Seiberg dualities of supersymmetric quiver gauge theories. Mathematically, this involves establishing mutations of quivers, which is in turn a variation on the theme of "learning to unknot". On the one hand, this leads us to a practical tool for establishing the computational complexity of different dualities. On the other hand, it also allows us to study how different network architectures learn how to trace Seiberg dualities. We find that for quivers with a modest number of quiver nodes (of order $10$), different network architectures consisting of transformers and multi-layer perceptrons tend to outperform deterministic algorithms. Supplementing the network by well-established pathfinder algorithms (essentially "Google Maps for quivers") leads to an additional improvement in the efficiency and accuracy of the search strategy. We anticipate that this class of questions can serve as a useful benchmark for frontier AI models applied to theoretical physics.