Relay, Don't Route: Adaptive Population Handoff for Cost-Efficient LLM-Driven Evolution

2026-08-06

AI Summary

This work addresses the inefficiency of existing large language model–based evolutionary search methods, which allocate computational resources at the level of individual model calls and neglect the stateful nature of evolutionary processes. To overcome this limitation, the authors propose a training-free, adaptive population handover framework that elevates budget allocation from single invocations to the population level. The approach leverages inexpensive models to concurrently explore multiple evolutionary trajectories and employs a multi-armed bandit scheduler to dynamically assess population-level gains. High-quality, diverse candidate sets are then opportunistically handed over to stronger models for refinement. Evaluated across four benchmark tasks under three budget regimes (12 experimental settings in total), the method achieves the highest average score in 11 settings, significantly outperforming current baselines.

Abstract

Large language model (LLM)-driven evolution has shown promise for program search and algorithm discovery, but relying on strong models throughout long evolutionary runs is costly. A natural alternative is to combine cheap and strong models under a fixed inference budget. However, existing approaches typically allocate models at the level of individual queries or mutation steps, overlooking that evolutionary search is \textit{stateful}: each generated candidate changes the population from which subsequent mutations are produced. We empirically analyze LLM-driven evolutionary trajectories and find that search progress is strongly front-loaded, early trajectory performance is informative but noisy, and cheap models recover much of the early progress achieved by strong models at lower cost. Motivated by these findings, we propose \textbf{\model}, a training-free framework that shifts budget allocation from individual calls to evolving populations through adaptive \textit{population handoff}. A cheap model explores multiple trajectories in short blocks allocated by a bandit scheduler. Relay Gain, defined as the marginal improvement of a compact, quality-diverse candidate bank constructed for handoff, serves as the scheduler reward and determines when to hand off. The curated candidates initialize a shared strong model population for refinement. Across four benchmarks and three budgets, \model achieves the highest mean score in 11 of 12 settings, outperforming competitive baselines. Our results suggest that in stateful search, budget allocation should be organized around the population, not the individual call.

Problem

• LLM-driven evolution
• cost-efficient inference
• stateful search
• population handoff
• evolutionary computation

Innovation

• population handoff
• Relay Gain
• LLM-driven evolution
• budget allocation
• stateful search