# Visual Words Meet BM25: Sparse Auto-Encoder Visual Word Scoring for Image Retrieval

arXiv: https://arxiv.org/abs/2603.05781 · HTML: https://mlsanity.com/paper/2603.05781/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 16 of its 28 references that are in the map.
- First posted on arXiv 6 Mar 2026.

## Builds on (most cited first, 16 of 16)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Dense Passage Retrieval for Open-Domain Question Answering](https://mlsanity.com/paper/2004.04906/) (2020)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [Billion-scale similarity search with GPUs](https://mlsanity.com/paper/1702.08734/) (2017)
- [Describing Textures in the Wild](https://mlsanity.com/paper/1311.3618/) (2013)
- [Fine-Grained Visual Classification of Aircraft](https://mlsanity.com/paper/1306.5151/) (2013)
- [SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features](https://mlsanity.com/paper/2502.14786/) (2025)
- [Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs](https://mlsanity.com/paper/1603.09320/) (2016)
- [SPLADE: Sparse Lexical and Expansion Model for First Stage Ranking](https://mlsanity.com/paper/2107.05720/) (2021)
- [Document Expansion by Query Prediction](https://mlsanity.com/paper/1904.08375/) (2019)
- [Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking](https://mlsanity.com/paper/1803.11285/) (2018)
- [A Few Brief Notes on DeepImpact, COIL, and a Conceptual Framework for Information Retrieval Techniques](https://mlsanity.com/paper/2106.14807/) (2021)
- [Context-Aware Sentence/Passage Term Importance Estimation For First Stage Retrieval](https://mlsanity.com/paper/1910.10687/) (2019)
- [Decoding Dense Embeddings: Sparse Autoencoders for Interpreting and Discretizing Dense Retrieval](https://mlsanity.com/paper/2506.00041/) (2025)
- [Learning Retrieval Models with Sparse Autoencoders](https://mlsanity.com/paper/2603.13277/) (2026)

## Built on by (most cited first, 2 of 2)

- [Latent Terms: Dense Retrievers Contain Trivially Extractable BM25-ready Zipfian Vocabularies](https://mlsanity.com/paper/2605.29384/) (2026)
- [REDI: Corpus Aware Patch Ranking for DINOv3 Token Reduction](https://mlsanity.com/paper/2606.31676/) (2026)

## Abstract (arXiv)

Dense image retrieval is accurate but offers limited interpretability and attribution, and it can be compute-intensive at scale. We present BM25-V, which applies Okapi BM25 scoring to sparse visual-word activations from a Sparse Auto-Encoder (SAE) on Vision Transformer patch features. Across a large gallery, visual-word document frequencies are highly imbalanced and follow a Zipfian-like distribution, making BM25's inverse document frequency (IDF) weighting well suited for suppressing ubiquitous, low-information words and emphasizing rare, discriminative ones. BM25-V retrieves high-recall candidates via sparse inverted-index operations and serves as an efficient first-stage retriever for dense reranking. Across seven benchmarks, BM25-V achieves Recall@200 ≥ 0.993, enabling a two-stage pipeline that reranks only K=200 candidates per query and recovers near-dense accuracy within 0.2% on average. An SAE trained once on ImageNet-1K transfers zero-shot to seven fine-grained benchmarks without fine-tuning, and BM25-V retrieval decisions are attributable to specific visual words with quantified IDF contributions.