A Structural Threshold in Decision Capacity Governs Collapse in Self-Play Reinforcement Learning

Fuente: arXiv

Saved in:

Bibliographic Details
Main Author:	Kujur, Arahan
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Artificial Intelligence 91A05, 91A26, 68T05, 68T42 I.2.6; I.2.8; F.2.2
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866914571084627968
author	Kujur, Arahan
author_facet	Kujur, Arahan
contents	We show that a threshold in decision capacity determines whether self-play reinforcement learning agents collapse under asymmetric rule perturbations. Across poker variants, matrix games, a dice game, and multiple learning algorithms, eliminating all positive-reach contingent decisions causes rapid convergence to a deterministic exploitation attractor, a fixed point at near-maximal loss. Preserving even a single positive-reach contingent decision point prevents this collapse. A frozen baseline and fixed-opponent control confirm that the mechanism is co-adaptation under constraint, not the perturbation itself. The phenomenon is timing-invariant, fully reversible upon action restoration, and intensifies under function approximation. These results establish a sharp threshold at zero reach-weighted contingent action capacity, with severity scaling continuously via reach-weighted capacity in the tested domains.
format	Preprint
id	arxiv_https___arxiv_org_abs_2605_16315
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	A Structural Threshold in Decision Capacity Governs Collapse in Self-Play Reinforcement Learning Kujur, Arahan Machine Learning Artificial Intelligence 91A05, 91A26, 68T05, 68T42 I.2.6; I.2.8; F.2.2 We show that a threshold in decision capacity determines whether self-play reinforcement learning agents collapse under asymmetric rule perturbations. Across poker variants, matrix games, a dice game, and multiple learning algorithms, eliminating all positive-reach contingent decisions causes rapid convergence to a deterministic exploitation attractor, a fixed point at near-maximal loss. Preserving even a single positive-reach contingent decision point prevents this collapse. A frozen baseline and fixed-opponent control confirm that the mechanism is co-adaptation under constraint, not the perturbation itself. The phenomenon is timing-invariant, fully reversible upon action restoration, and intensifies under function approximation. These results establish a sharp threshold at zero reach-weighted contingent action capacity, with severity scaling continuously via reach-weighted capacity in the tested domains.
title	A Structural Threshold in Decision Capacity Governs Collapse in Self-Play Reinforcement Learning
topic	Machine Learning Artificial Intelligence 91A05, 91A26, 68T05, 68T42 I.2.6; I.2.8; F.2.2
url	https://arxiv.org/abs/2605.16315