Abstract
Self-improving learning systems adapt data selection, optimization, and auxiliary symbolic components, inducing nonstationary objectives outside standard learning assumptions. We introduce \textsc{PRAXIS}, a co-evolutionary framework that models generators, learners, and symbolic archives as interacting dynamical processes. We prove that KL-constrained generator updates and controlled archive-weight movement bound one-step objective drift, that archive updates suppress a program relative to any fixed comparator with a persistent cumulative utility advantage under sub-Gaussian noise, and that stochastic gradient descent achieves an average-stationarity guarantee whose degradation is governed by cumulative objective drift. Experiments across visual robustness, relational graph reasoning, and algorithmic graph reasoning exhibit generator stabilization, decreasing learner loss, and archive concentration consistent with these theoretical mechanisms.
Keywords
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Margapuri, V., & Teber, M. (2026). PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives. https://omanscience.com/en/articles/praxis-learning-dynamics-of-self-improving-models-with-symbolic-archives
MLA 9
Margapuri, Venkat, and Mustafa Teber. "PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives." https://omanscience.com/en/articles/praxis-learning-dynamics-of-self-improving-models-with-symbolic-archives.
Chicago (author–date)
Margapuri, Venkat, and Mustafa Teber. 2026. "PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives." https://omanscience.com/en/articles/praxis-learning-dynamics-of-self-improving-models-with-symbolic-archives.
Harvard
Margapuri, V. and Teber, M. (2026) 'PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives', Available at: https://omanscience.com/en/articles/praxis-learning-dynamics-of-self-improving-models-with-symbolic-archives.
Vancouver
Margapuri V, Teber M. PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives. https://omanscience.com/en/articles/praxis-learning-dynamics-of-self-improving-models-with-symbolic-archives
IEEE
V. Margapuri, and M. Teber, "PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives," https://omanscience.com/en/articles/praxis-learning-dynamics-of-self-improving-models-with-symbolic-archives.