A Reproducibility Study of PLAID

TL;DR

This study reproduces and analyzes the PLAID algorithm for document retrieval, comparing it with re-ranking baselines, and explores how parameter tuning and recent modifications affect efficiency and effectiveness trade-offs.

Contribution

It fills gaps in the original PLAID work, evaluates its parameters, compares it with re-ranking baselines, and analyzes token clusters to understand retrieval performance.

Findings

PLAID's Pareto frontier depends on three key parameters.

Re-ranking with ColBERTv2 on BM25 pools offers better efficiency-effectiveness trade-offs at low latency.

Recent neighbor-based re-ranking modifications improve performance across all operational points.

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 then compare PLAID with an important baseline missing from the paper: re-ranking a lexical system. We find that applying ColBERTv2 as a re-ranker atop an initial pool of BM25 results provides better efficiency-effectiveness trade-offs in low-latency settings. However, re-ranking cannot reach peak effectiveness at higher latency settings due to limitations in recall of lexical matching and provides a poor approximation of an exhaustive ColBERTv2 search. We find that recently proposed modifications to re-ranking that pull in the neighbors of top-scoring documents overcome this limitation, providing a Pareto frontier across all operational points for ColBERTv2 when evaluated using a well-annotated dataset. Curious about why re-ranking methods are highly competitive with PLAID, we analyze the token representation clusters PLAID uses for retrieval and find that most clusters are predominantly aligned with a single token and vice versa. Given the competitive trade-offs that re-ranking baselines exhibit, this work highlights the importance of carefully selecting pertinent baselines when evaluating the efficiency of retrieval engines.