用进化算法让大模型学会类人推理，弱模型也能变强。
Evolutionary System 2 Reasoning: An Empirical Proof
- 通过多模型进化筛选，优化推理能力而非特定技能。
- 仅用简单循环进化，70亿参数模型可超越GPT-5的推理表现。
- 适合对通用智能、模型演化感兴趣的读者。
机器智能是使机器具备类人智能的终极目标。尽管大语言模型（LLMs）在众多下游任务中展现出强大专有技能，但在通用智能方面仍显不足。鉴于智力与系统2推理（慢思考）之间的相关性，本文提出一个关键问题：大模型能否像人类一样被演化出推理能力（而非特定技能）？为此，我们提出进化推理优化（ERO）框架，通过对一组大模型进行“适者生存”式进化，寻找具备强推理能力的个体。针对具体推理任务，ERO首先初始化多个大模型作为种群，再通过进化策略提升种群中最佳个体的量化推理得分。在代表性测试集上的实验揭示了两个令人意外的发现：其一，最新的大模型如GPT-5仍表现出有限的系统2推理能力；其二，仅通过简单的ERO进化循环，一个相对弱小的模型（Qwen-7B）即可涌现出强大的推理能力。项目代码已开源，可访问 https://github.com/MetaEvo/ERO 复现。
原文摘要 · Abstract (English)
Machine intelligence marks the ultimate dream of making machines' intelligence comparable to human beings. While recent progress in Large Language Models (LLMs) show substantial specific skills for a wide array of downstream tasks, they more or less fall shorts in general intelligence. Following correlation between intelligence and system 2 reasoning (slow thinking), in this paper, we aim to answering a worthwhile research question: could machine intelligence such as LLMs be evolved to acquire reasoning ability (not specific skill) just like our human beings? To this end, we propose evolutionary reasoning optimization (ERO) framework which performs survival of the fittest over a population of LLMs to search for individual with strong reasoning ability. Given a reasoning task, ERO first initializes multiple LLMs as a population, after which an evolutionary strategy evolves the population to maximize quantified reasoning score of the best individual. Based on experiments on representative testsuites, we claim two surprising empirical discoveries: i) the latest LLMs such as GPT-5 still show limited system 2 reasoning ability; ii) with simple evolution-loop of ERO, a relatively weak model (Qwen-7B) could be enhanced to emerge powerful reasoning ability. Our project can be accessed at https://github.com/MetaEvo/ERO for reproduction needs.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。