Evolving Parallel Algorithm Portfolios via Potential-Aware Instance Generation with LLMs

This paper proposes PIAC, a potential-aware co-evolution framework that enhances LLM-based algorithm portfolios' few-shot generalization without reference solutions, outperforming state-of-the-art baselines on TSP and CVRP.

Cooperative Coevolution for Resource-Constrained Agentic LLM Post-Training

CoPES accelerates memory-efficient, full-parameter agentic LLM post-training on limited GPUs via cooperative coevolution.

MEGO: Learning Mixture-of-Experts for General-Purpose Binary Optimization

This article introduces MEGO, a novel general-purpose neural optimizer for binary optimization under the black-box setting, intended for broad applicability across diverse binary optimization problem classes with minimal problem-specific customization.

Evolving Generalizable Parallel Algorithm Portfolios via Domain-Agnostic Instance Generation

高斯过程：从随机过程中来，到机器学习中去

高斯过程是一种经典、拥有良好性质的非参模型，其本身可以看作一个随机过程建模，由于其非参概率模型的特性，在一些低维度少样本的连续参数优化问题上，高斯过程模型有着非参广泛的应用。

基于双重argsort的样本位次便捷计算

两次argsort就能直接拿到样本位次？靠谱吗，性能咋样？