SOURCE-LINKED INTELLIGENCE

EigenLI: Spectral Approximations to Late Interaction

arXiv · AI, language, vision and robotics · article · Sep 7, 2026 · UTC

Late-interaction models such as ColBERT achieve strong effectiveness by representing each document with many token-level vectors, but this expressivity leads to large indexing cost, storage footprints and expensive MaxSim scoring. We show that late-interaction representations exhibit an intrinsic low-rank structure: document token embeddings concentrate in a low-dimensional subspace that preserves most of the retrieval signal. Leveraging this observation, we introduce EigenLI, a spectral approximation framework that compresses late-interaction representations via document-specific low-dimensio

Read original source ↗ Open in workspace

recordType

paper

region

Global