MECo Framework Enables Zero-Shot Cross-Problem Combinatorial Optimization

MECo uses a multi-task evolutionary framework to generate and recombine task-conditioned heuristics. This approach achieves zero-shot cross-problem generalization by using a transfer gap based on cross-task population performance to guide heuristic interaction.