REMAG is a specialized metagenomic binning tool designed for recovering high-quality eukaryotic genomes from mixed prokaryotic-eukaryotic samples. It uses contrastive learning with Siamese neural networks to generate meaningful contig embeddings for clustering.