AdaptEvolve: Improving Efficiency of Evolutionary AI Agents through Adaptive Model Selection

2026-02-12

AI Summary

This work addresses the challenge of balancing computational cost and inference performance in evolutionary AI agents, where frequent calls to large language models (LLMs) incur substantial overhead. To this end, the authors propose an adaptive LLM selection mechanism based on generation confidence, which dynamically routes tasks within an evolutionary sequence optimization framework to the most capable yet efficient model by real-time assessment of task solvability. This approach is the first to leverage intrinsic generation confidence for online model routing—without relying on static heuristics or external controllers—and explicitly models and exploits model uncertainty. Experiments demonstrate that the method reduces inference costs by 37.9% on average across multiple benchmarks while retaining 97.5% of the accuracy achieved by a static, high-capacity LLM, thereby significantly advancing the Pareto frontier of efficiency and performance.

Technology Category

• Machine Learning: Evolutionary Learning
• Search and Optimization: Evolutionary Computation
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Evolutionary agentic systems intensify the trade-off between computational efficiency and reasoning capability by repeatedly invoking large language models (LLMs) during inference. This setting raises a central question: how can an agent dynamically select an LLM that is sufficiently capable for the current generation step while remaining computationally efficient? While model cascades offer a practical mechanism for balancing this trade-off, existing routing strategies typically rely on static heuristics or external controllers and do not explicitly account for model uncertainty. We introduce AdaptEvolve: Adaptive LLM Selection for Multi-LLM Evolutionary Refinement within an evolutionary sequential refinement framework that leverages intrinsic generation confidence to estimate real-time solvability. Empirical results show that confidence-driven selection yields a favourable Pareto frontier, reducing total inference cost by an average of 37.9% across benchmarks while retaining 97.5% of the upper-bound accuracy of static large-model baselines. Our code is available at https://github.com/raypretam/adaptive_llm_selection.

Problem

• evolutionary AI agents
• large language models
• computational efficiency
• reasoning capability
• model selection

Innovation

• adaptive model selection
• evolutionary AI agents
• intrinsic confidence
• multi-LLM routing
• computational efficiency