PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives

Key points

• Self-improving learning systems adapt data selection, optimization, and auxiliary symbolic components, inducing nonstationary objectives outside standard learning assumptions.1
• 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.1
• 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.1

Sources (1)

• [1]PRAXIS: Learning Dynamics of Self-Improving Models with Symbolic Archives
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 12:11 PM

  We introduce PRAXIS, a co-evolutionary framework that models generators, learners, and symbolic archives as interacting dynamical processes.

  Self-improving learning systems adapt data selection, optimization, and auxiliary symbolic components, inducing nonstationary objectives outside standard learning assumptions.

Extractive summary: sentences quoted from the sources.