AI Paper Atlas

Research map / Learning theory and algorithms / Combinatorial and search optimization

DESSERT: An Efficient Algorithm for Vector Set Search with Vector Set Queries

Joshua Engels, Benjamin Coleman, Vihan Lakshman, Anshumali Shrivastava

NeurIPS 2023 · Learning theory and algorithms › Combinatorial and search optimization · approach: Heuristic and local search algorithms · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use DESSERT instead of PLAID for k=10 retrievalColBERT on MS MARCO development · MS MARCO development, k=10latency: 45.1 -> 9.5Improved
Use DESSERT instead of PLAID for k=10 retrievalColBERT on MS MARCO development · MS MARCO development, k=10MRR@10: 39.2 ± 1.15 -> 35.7 ± 1.14Hurt
Use DESSERT instead of PLAID for k=10 retrieval at the higher-latency pointColBERT on MS MARCO development · MS MARCO development, k=10MRR@10: 39.2 ± 1.15 -> 37.2 ± 1.14Hurt
Use DESSERT instead of PLAID for k=10 retrieval at the higher-latency pointColBERT on MS MARCO development · MS MARCO development, k=10latency: 45.1 -> 15.5Improved
Use DESSERT instead of PLAID for k=1000 retrievalColBERT on MS MARCO development · MS MARCO development, k=1000latency: 100 -> 22.7Improved
Use DESSERT instead of PLAID for k=1000 retrievalColBERT on MS MARCO development · MS MARCO development, k=1000R@1000: 97.5 ± 0.36 -> 95.1 ± 0.49Hurt
Use DESSERT instead of PLAID for k=1000 retrieval at the higher-latency pointColBERT on MS MARCO development · MS MARCO development, k=1000latency: 100 -> 32.3Improved
Use DESSERT instead of PLAID for k=1000 retrieval at the higher-latency pointColBERT on MS MARCO development · MS MARCO development, k=1000R@1000: 97.5 ± 0.36 -> 96.0 ± 0.45Hurt
Use DESSERT rather than optimized PyTorch brute forceVector set search on synthetic GloVe sets · Synthetic GloVe sets, N=1000speedup: 10-50xImproved
Use DESSERT rather than PLAIDColBERT passage retrieval · LoTTE, Lifestyle, Technology, Recreation, and Science splitslatency: 2-5xImproved
Use single-vector ScaNN search instead of vector-set retrievalPassage retrieval · Passage retrieval, corpus unspecifiedRecall@1000: 0.77Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy