A Reproducibility Study of PLAID

arXiv:2404.14989v1 Announce Type: cross
Abstract: The PLAID (Performance-optimized Late Interaction Driver) algorithm for ColBERTv2 uses clustered term representations to retrieve and progressively prune documents for final (exact) document scoring. In this paper, we reproduce and fill in missing gaps from the original work. By studying the parameters PLAID introduces, we find that its Pareto frontier is formed of a careful balance among its three parameters; deviations beyond the suggested settings can substantially increase latency without necessarily improving its effectiveness. We …

abstract algorithm arxiv cs.cl cs.ir document documents driver paper parameters pareto performance plaid reproducibility scoring study studying type work