After the party comes the free lunch: regularizing ColBERT models to enhance pooling capabilities and reduce index footprint

Back to Articles

Hierarchical pooling already lets you cut ColBERT storage in half without hurting retrieval. We now show that STE-based regularization, the same trick we used to fix MUVERA/SMVE, pushes this much further: 99.4% retention at 5× compression, without hurting full-token performance. The key insight is that the same as for MUVERA/SMVE: the best way to achieve compressability is to directly train for it.

Table of Contents

Hierarchical Pooling: The Quiet Workhorse

A Regularization That Keeps on Giving