Behavior of DCO methods on dynamic data

Investigate the behavior of Distance Comparison Operation (DCO) methods when operating on dynamic data, including scenarios where data are incrementally inserted or otherwise updated.

Background

Most existing evaluations of DCO methods focus on static datasets, whereas many practical vector database deployments require frequent insertions or updates. The paper explicitly notes that, due to limited prior evaluations, how DCO methods behave in such dynamic settings was previously unknown.

References

While DCO methods have demonstrated strong performance on certain static datasets, their behavior on dynamic data remains unknown due to limited evaluations on this scenario.

— Distance Comparison Operations Are Not Silver Bullets in Vector Similarity Search: A Benchmark Study on Their Merits and Limits  (2604.02801 - Zheng et al., 3 Apr 2026) in Section 'Robustness on Dynamic Data'