The paper bioMoR: Biology-Guided Mixture-of-Recursions for Effective Genomic Learning proposes a novel framework for genomic learning, called bioMoR. By applying Mixture-of-Recursions (MoR) to gene-level and pathway-level learning, bioMoR improves the efficiency of high-dimensional omics analysis. The main contributions of bioMoR include three aspects: graph-based information sharing, structural bias guiding self-attention, and graph-aware routing. These techniques are centered on the insight that additional knowledge of token interaction can effectively help models construct embeddings and select which tokens should be learned more deeply. The results show that bioMoR improves average macro-F1 by 8.2 percentage points and balanced accuracy by 7.1 percentage points over the strongest biology-agnostic MoR baseline, using 75% fewer parameters and up to 58% fewer FLOPs than a non-recursive Transformer. Blogger's Review: bioMoR is a promising framework for genomic learning that significantly improves model efficiency and accuracy by integrating structured biological knowledge. It also provides biological interpretability and token-specific recursion depths, enabling researchers to better understand the relationships between genes and pathways.