TensorFlow Ranking is a specialized library for learning-to-rank problems built on TensorFlow. It provides commonly used loss functions (pointwise, pairwise, listwise), ranking metrics (MRR, NDCG), and scoring functions including LambdaLoss for direct ranking metric optimization. The library also supports unbiased learning-to-rank from biased feedback data and includes tools for handling sparse features and embeddings.