BACKGROUND: Dissecting the genetic basis of cell-type-specific gene regulation during gametogenesis remains a major challenge. Population-level studies mask cellular heterogeneity, while the extreme data sparsity of single-cell profiling precludes robust genetic mapping. RESULTS: We present GAMETE (Genetic Architecture Mapping of Epigenetic Traits at single-nucleus resolution), an integrated framework combining MPS-ATAC-seq, a high-throughput single-nucleus ATAC-seq method, with robust computational approaches for genotype inference and genetic mapping from sparse single-nucleus data. Applying GAMETE to hybrid rice pollen, we simultaneously profile chromatin accessibility and infer the genotypes of 4,887 individual haploid nuclei, revealing distinct chromatin landscapes across pollen cell types. This uncovers an unexpected elevation of chromatin accessibility at gypsy retrotransposons in microspores preceding their silencing in sperm cells, exposing a surveillance gap that may contribute to genome expansion. The synchronized profiling enables us to construct a single-cell recombination map, map segregation distortion loci and trace their developmental origins, and identify 16,113 chromatin accessibility QTLs (caQTLs), 48.6% of which exhibit pronounced cell-type specificity. Analysis of introgression lines supports a cell-type-specific trans-acting caQTL regulating the pollen-essential gene DTM1. CONCLUSIONS: This work establishes GAMETE as a broadly applicable framework for dissecting the genetic architecture of chromatin accessibility during gametogenesis without the need for subsequent generations.