LoongFlow: Directed Evolutionary Search via a Cognitive Plan-Execute-Summarize Paradigm

TL;DR

LoongFlow is a novel self-evolving agent framework that leverages a cognitive Plan-Execute-Summarize paradigm with hybrid memory and evolutionary strategies to improve solution quality and efficiency in complex search spaces.

Contribution

It introduces LoongFlow, integrating LLMs into a reasoning-based evolutionary paradigm with hybrid memory, achieving state-of-the-art results with lower computational costs.

Findings

Outperforms baselines by up to 60% in efficiency.

Achieves superior solutions on benchmark and Kaggle tasks.

Effectively maintains diversity and avoids stagnation.

Abstract

The transition from static Large Language Models (LLMs) to self-improving agents is hindered by the lack of structured reasoning in traditional evolutionary approaches. Existing methods often struggle with premature convergence and inefficient exploration in high-dimensional code spaces. To address these challenges, we introduce LoongFlow, a self-evolving agent framework that achieves state-of-the-art solution quality with significantly reduced computational costs. Unlike "blind" mutation operators, LoongFlow integrates LLMs into a cognitive "Plan-Execute-Summarize" (PES) paradigm, effectively mapping the evolutionary search to a reasoning-heavy process. To sustain long-term architectural coherence, we incorporate a hybrid evolutionary memory system. By synergizing Multi-Island models with MAP-Elites and adaptive Boltzmann selection, this system theoretically balances the exploration-exploitation trade-off, maintaining diverse behavioral niches to prevent optimization stagnation. We instantiate LoongFlow with a General Agent for algorithmic discovery and an ML Agent for pipeline optimization. Extensive evaluations on the AlphaEvolve benchmark and Kaggle competitions demonstrate that LoongFlow outperforms leading baselines (e.g., OpenEvolve, ShinkaEvolve) by up to 60% in evolutionary efficiency while discovering superior solutions. LoongFlow marks a substantial step forward in autonomous scientific discovery, enabling the generation of expert-level solutions with reduced computational overhead.