AnchorFold: A Focus-Then-Fold Framework via Recursive Attention Propagation for Efficient Multi-Vector Visual Document Retrieval

arXiv:2608.08732 · cs.CV, cs.CL, cs.IR · Submitted 2026-08-09 · Read on arXiv

Haoyu Zuo, Yibo Yan, Xin Zou, Shuliang Liu, Yi Cao, Mingdong Ou, Xuming Hu

cs.CV, cs.CL, cs.IR

Submitted: 2026-08-09

Updated: 2026-08-11

Comments: 24 pages, 7 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 49× in this paper · explained in 2 episodes
• Calibration — 31× in this paper · explained in 9 episodes
• RAG — 27× in this paper · explained in 3 episodes
• Pruning — 26× in this paper · explained in 4 episodes
• Embeddings — 16× in this paper · explained in 2 episodes
• Clustering — 15× in this paper · explained in 2 episodes
• Visual Document Retrieval — 13× in this paper
• Fidelity — 11× in this paper · explained in 2 episodes
• Semantic Clustering — 10× in this paper · explained in 2 episodes
• Generalization — 4× in this paper · explained in 14 episodes
• Ablation — 3× in this paper · explained in 1 episode
• Tokenization — 3× in this paper · explained in 2 episodes
• Ablation Study — 2× in this paper · explained in 3 episodes
• Adam — 2× in this paper · explained in 3 episodes
• Evaluation Protocol — 2× in this paper · explained in 2 episodes
• Scoring Function — 2× in this paper · explained in 2 episodes
• Training-Free Framework — 2× in this paper · explained in 2 episodes
• Transformer — 2× in this paper · explained in 2 episodes
• Vector Representations — 2× in this paper · explained in 2 episodes
• Cosine Similarity — 1× in this paper · explained in 3 episodes
• Latent Space — 1× in this paper · explained in 7 episodes
• Papers — 1× in this paper · explained in 3 episodes
• Performance Gap — 1× in this paper · explained in 2 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Robustness — 1× in this paper · explained in 6 episodes
• Score Retention — 1× in this paper
• Transformers — 1× in this paper · explained in 2 episodes

Sources

• Qwen2.5-VL Technical Report
• ReinPool: Reinforcement Learning Pooling Multi-Vector Embeddings for Retrieval System
• ColBERT-Zero: To Pre-train Or Not To Pre-train ColBERT models
• Reducing the Footprint of Multi-Vector Retrieval with Minimal Performance Impact via Token Pooling
• Nemotron ColEmbed V2: Top-Performing Late Interaction Embedding Models for Visual Document Retrieval
• jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval
• CausalEmbed: Auto-Regressive Multi-Vector Generation in Latent Space for Visual Document Embedding
• A Brief Comparison of Training-Free Multi-Vector Sequence Compression Methods
• LEMUR: Learned Multi-Vector Retrieval
• Structural Anchor Pruning: Training-Free Multi-Vector Compression for Visual Document Retrieval
• Towards Storage-Efficient Visual Document Retrieval: An Empirical Study on Reducing Patch-Level Embeddings
• ViDoRe Benchmark V2: Raising the Bar for Visual Retrieval
• Multi-Vector Index Compression in Any Modality
• CRISP: Clustering Multi-Vector Representations for Denoising and Pruning
• MM-Matryoshka: Towards Budget-Elastic Visual Document Retrieval via a 2D Multimodal Matryoshka Training Framework
• Unlocking Multimodal Document Intelligence: From Current Triumphs to Future Frontiers of Visual Document Retrieval
• Visual Late Chunking: An Empirical Study of Contextual Chunking for Efficient Visual Document Retrieval
• Beyond the Grid: Layout-Informed Multi-Vector Retrieval with Parsed Visual Document Representations
• DocPruner: A Storage-Efficient Framework for Multi-Vector Visual Document Retrieval via Adaptive Patch-Level Embedding Pruning

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models