Dynamic Superblock Pruning for Fast Learned Sparse Retrieval
About
This paper proposes superblock pruning (SP) during top-k online document retrieval for learned sparse representations. SP structures the sparse index as a set of superblocks on a sequence of document blocks and conducts a superblock-level selection to decide if some superblocks can be pruned before visiting their child blocks. SP generalizes the previous flat block or cluster-based pruning, allowing the early detection of groups of documents that cannot or are less likely to appear in the final top-k list. SP can accelerate sparse retrieval in a rank-safe or approximate manner under a high-relevance competitiveness constraint. Our experiments show that the proposed scheme significantly outperforms state-of-the-art baselines on MS MARCO passages on a single-threaded CPU.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Information Retrieval | SCIDOCS | nDCG@1015.9 | 24 | |
| Information Retrieval | NQ BEIR | -- | 20 | |
| Information Retrieval | DBPedia BEIR (test) | nDCG@1043.6 | 18 | |
| Information Retrieval | FIQA BEIR (test) | nDCG@1035.5 | 15 | |
| Information Retrieval | Arguana BEIR | nDCG48.7 | 11 | |
| Information Retrieval | Quora BEIR | nDCG83.2 | 11 | |
| Information Retrieval | C-FEVER BEIR | nDCG24.5 | 5 | |
| Information Retrieval | Touche BEIR | nDCG0.273 | 5 | |
| Information Retrieval | FEVER BEIR | nDCG0.795 | 5 | |
| Information Retrieval | Hotpot BEIR | nDCG0.685 | 5 |