Structural Anchor Pruning: Training-Free Multi-Vector Compression for Visual Document Retrieval

TL;DR

Structural Anchor Pruning (SAP) is a training-free, query-agnostic framework that effectively compresses visual document retrieval models by pruning over 90% of tokens while retaining high retrieval quality.

Contribution

SAP introduces a novel self-calibrating, training-free pruning method with layer-wise diagnostics and structural analysis, enabling high compression without per-model tuning.

Findings

SAP retains over 90% of NDCG@5 after pruning more than 90% of tokens.

SAP outperforms existing methods that rely on heuristics or training.

Analysis reveals a stable 'Structural Plateau' in visual representations within the backbone.

Abstract

Recent Vision-Language Models (e.g., ColPali) enable fine-grained Visual Document Retrieval (VDR) but incur prohibitive multi-vector index storage overhead. Existing training-free pruning methods either rely on heuristic layer choices or degrade sharply under aggressive compression, leading prior work to argue that effective high-compression pruning requires query-dependent training. We challenge this view with Structural Anchor Pruning (SAP), a self-calibrating, training-free, and query-agnostic index-time pruning framework with three components: (i) Score Retention (SR), a white-box per-layer compression diagnostic; (ii) SR-guided window selection, a procedure that automatically locates the structural pruning region for any backbone with no per-model hyperparameters; and (iii) a visual in-degree centrality scorer that identifies anchor patches within the selected window. On the ViDoRe v1/v2 benchmarks across three architectures spanning 18, 28, and 36 backbone layers, SAP retains over 90\% of NDCG@5 while pruning more than 90\% of visual tokens, without any per-model parameter tuning. Our layer-resolved SR analysis reveals an Alignment-Aggregation Divergence: the document's visual structure is preserved as a stable ``Structural Plateau'' within the backbone, but the final layers reshape this representation into a sparse, query-aligned form that is no longer suitable for pruning. This is the mechanistic reason SAP succeeds where final-layer methods fail.