Cost-Penalized Fitness in FMA-Orchestrated Mixture of Experts: Experimental Evidence for Molecular Memory in Domain Adaptation

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autor principal: Jaraiz, Martin
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918423592697856
author Jaraiz, Martin
author_facet Jaraiz, Martin
contents We present experimental results from seven controlled runs of nanoFMT, a Free-Market Algorithm (FMA) orchestrated transformer with dynamic Mixture-of-Experts (MoE) management. The experiments address a fundamental question for advanced LLM development: how should an MoE system manage its expert pool when operating at full capacity under changing data distributions? We demonstrate that cost-penalized fitness metrics, combined with a linear grace period for newborn experts, produce a system that accumulates domain expertise through diversification rather than replacement. The central result is a round-trip domain shift experiment showing 9-11x faster recovery when returning to a previously learned domain, with zero expert births or replacements required. This "molecular memory" effect -- where dormant experts survive and reactivate when their domain returns -- has no analogue in current MoE management approaches. A preliminary cost analysis estimates annual savings of $39.1M and 27.1 GWh energy reduction for an OpenAI-scale provider under a moderate scenario.
format Preprint
id arxiv_https___arxiv_org_abs_2604_00812
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Cost-Penalized Fitness in FMA-Orchestrated Mixture of Experts: Experimental Evidence for Molecular Memory in Domain Adaptation
Jaraiz, Martin
Machine Learning
We present experimental results from seven controlled runs of nanoFMT, a Free-Market Algorithm (FMA) orchestrated transformer with dynamic Mixture-of-Experts (MoE) management. The experiments address a fundamental question for advanced LLM development: how should an MoE system manage its expert pool when operating at full capacity under changing data distributions? We demonstrate that cost-penalized fitness metrics, combined with a linear grace period for newborn experts, produce a system that accumulates domain expertise through diversification rather than replacement. The central result is a round-trip domain shift experiment showing 9-11x faster recovery when returning to a previously learned domain, with zero expert births or replacements required. This "molecular memory" effect -- where dormant experts survive and reactivate when their domain returns -- has no analogue in current MoE management approaches. A preliminary cost analysis estimates annual savings of $39.1M and 27.1 GWh energy reduction for an OpenAI-scale provider under a moderate scenario.
title Cost-Penalized Fitness in FMA-Orchestrated Mixture of Experts: Experimental Evidence for Molecular Memory in Domain Adaptation
topic Machine Learning
url https://arxiv.org/abs/2604.00812