) or not using () supervised/unsupervised data. Reporting accuracy @ k. System Sup. Unsup. A@20 A@100 BM25 - - 62.9 78.1 SEAL 76.2 86.3 (LM+FM74.8 85.4 intersective) 61.7 76.3

Table 10: Retrieval results on the KIL T test set(s). Reporting page-level R-precision (higher is better). Results are taken from the eval.ai Table 11: KIL T scores on the KIL T test set(s). Results are taken from the eval.ai