EvoSkillRec: Skill-Genome Evolution for Recommender Architecture Discovery

2026-09-28

AI Summary

This study addresses the challenges of strong human bias, fragile LLM-driven code evolution, and poor innovation reusability in recommender system architecture design by proposing a Skill Genome Evolution framework. This method introduces a pioneering skill-code dual-coupled evolutionary space that decomposes models into atomic skills, leveraging LLMs to drive code synthesis and typed mutation-recombination within constrained spaces for novel module generation. Furthermore, it establishes a skill library to validate and reuse effective innovations, while an automated research controller adaptively allocates exploration budgets to achieve cumulative architectural discovery. Experimental results demonstrate that this framework significantly enhances performance in CTR prediction, multi-task learning, and multi-domain learning, while also optimizing computational efficiency for generative ranking models.

Abstract

Modern recommender systems advance not only by scaling data and parameters, but also by encoding task-specific inductive biases through architecture, including sparse feature interactions for click-through rate (CTR) prediction, temporal attention for sequential recommendation, and expert routing for multi-task learning. However, these biases are typically human expert designed or searched within predefined operator spaces. Although Recent LLM-driven code evolution expands this space, unconstrained edits often produce invalid or ineffective architectures, underuse established architecture design knowledge, and fail to preserve successful innovations for reuse. We introduce EvoSkillRec, a promotion-and-reuse framework for cumulative recommender architecture evolution. It first decomposes recommenders into atomic executable skills and represents architectures as typed skill genomes, with each skill equipped with input--output types, semantic annotations, and implementation code. We then evolve models with different tasks through two coupled spaces: a constrained skill--space that mutates, recombines, specializes, and reuses validated skills, and an open-ended code--space in which LLM planners and synthesizers invent new skill modules using prior evolution traces and accumulated experience. An autoresearch controller evaluates candidates, diagnoses failures, retrieves relevant skills, promotes validated innovations into the skill library, and adaptively allocates the proposal budget between the two spaces. Extensive experiments on CTR prediction, multi-task learning, and multi-domain learning, including resource-constrained co-optimization of predictive quality and model FLOPs utilization in generative ranking models, consistently demonstrate the effectiveness of our proposed EvoSkillRec.

Problem

• Recommender Systems
• Architecture Search
• Neural Architecture Discovery
• Code Evolution

Innovation

• Architecture Search
• Skill Genome
• LLM-driven Evolution
• Recommender Systems
• AutoResearch