ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
The DOI matches its Crossref record; 24 of the 24 references it cites were found in a registry or an index.
Abstract
Recent progress in Natural Language Understanding (NLU) is driving fast-paced advances in Information Retrieval (IR), largely owed to fine-tuning deep language models (LMs) for document ranking. While remarkably effective, the ranking models based on these LMs increase computational cost by orders of magnitude over prior approaches, particularly as they must feed each query-document pair through a massive neural network to compute a single relevance score. To tackle this, we present ColBERT, a novel ranking model that adapts deep LMs (in particular, BERT) for efficient retrieval. ColBERT introduces a late interaction architecture that independently encodes the query and the document using BERT and then employs a cheap yet powerful interaction step that models their fine-grained similarity. By delaying and yet retaining this fine-granular interaction, ColBERT can leverage the expressiveness of deep LMs while simultaneously gaining the ability to pre-compute document representations offline, considerably speeding up query processing. Crucially, ColBERT's pruning-friendly interaction mechanism enables leveraging vector-similarity indexes for end-to-end retrieval directly from millions of documents. We extensively evaluate ColBERT using two recent passage search datasets. Results show that ColBERT's effectiveness is competitive with existing BERT-based models (and outperforms every non-BERT baseline), while executing two orders-of-magnitude faster and requiring up to four orders-of-magnitude fewer FLOPs per query.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“The only ML component is a CrossEncoder reranker ( mxbai-rerank-large-v1 , 435M parameters, DeBERTaV3) fused with ColBERT (Khattab & Zaharia, 2020) via Reciprocal Rank Fusion (RRF), applied after retrieval is complete to prioritize passages before token-budget truncation.”
From SmartSearch: How Ranking Beats Structure for Conversational Memory Retrieval (2026) · cites this paper as Khattab & Zaharia, 2020
In the citing paper
“While dense retrievers like Dense Passage Retrieval (DPR) [14] and late-interaction models like ColBERT [15] improve recall on single-hop queries, they often struggle with multi-hop reasoning, where the answer depends on composing information from multiple disjoint documents [20, 6].”
From Replace, Don't Expand: Mitigating Context Dilution in Multi-Hop RAG via Fixed-Budget Evidence Assembly (2025) · cites this paper as [15]
In the citing paper
“Research on retrieval optimization has explored multiple techniques, yet studies specifically focused on prefetching size in RaLM systems are still limited.”
From Exploring Optimal Prefetching Sizes in RaLMSpec to Enhance Retrieval-Augmented Generation Efficiency (2025) · cites this paper as [9]
In the citing paper
“Early RAG implementations, however, often relied on document retrieval methods that introduced noise and excessive context [26, 55, 60, 122].”
From LEGO-GraphRAG: Modularizing Graph-Based Retrieval-Augmented Generation for Design Space Exploration (2025) · cites this paper as [55]
In the citing paper
“Standard retrieval methods like RAG and MemoryBank (Zhong et al., 2024) rely fundamentally on vector similarity (Karpukhin et al., 2020; Khattab and Zaharia, 2020), representing memories as isolated points in embedding space (Hu et al., 2025). Consequently, they struggle with queries requiring causal bridging between semantically dissimilar or distant events (Yang et al., 2018; Qi et al., 2019; Tr”
From Synapse: Empowering LLM Agents with Episodic-Semantic Memory via Spreading Activation (2026) · cites this paper as Khattab and Zaharia, 2020
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404 · cited 1×
“At a high level, TK improves the KNRM architecture described earlier: while KNRM employs kernel pooling on top of word-embedding-based interaction, TK uses a Transformer [34] component for contextually encoding queries and documents before kernel pooling.”
MethodBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173 · cited 2×
“Among them, a recent approach has emerged that ne-tunes deep pre-trained language models (LMs) like ELMo [29] and BERT [5] for estimating relevance.”
BackgroundGoogle's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
Wu, Y., Schuster, M., Chen, Z., Le, Q. V., Norouzi, M., Macherey, W., Krikun, M., Cao, Y., Gao, Q., Macherey, K., et al. (2016) · arXiv · doi:10.48550/arxiv.1609.08144 · Cited by 5,319
Billion-scale similarity search with GPUs
Johnson, J., Douze, M., & Jégou, H. (2017) · arXiv · doi:10.48550/arxiv.1702.08734 · Cited by 4,140 · cited 2×
“Additionally, it enables ColBERT to leverage vector-similarity search indexes (e.g., [1, 16]) to retrieve the top-k results directly from a large document collection, substantially improving recall over models that only re-rank the output of term-based retrieval.”
MethodMS MARCO: A Human Generated MAchine Reading COmprehension Dataset
Bajaj, P., Campos, D., Craswell, N., Deng, L., Gao, J., Liu, X., Majumder, R., McNamara, A., Mitra, B., Nguyen, T., et al. (2016) · arXiv · doi:10.48550/arxiv.1611.09268 · Cited by 2,473
Robertson, S. E., Walker, S., Jones, S., Hancock‐Beaulieu, M., & Gatford, M. (1994) · Text REtrieval Conference · Cited by 2,212
Learning deep structured semantic models for web search using clickthrough data
Huang, P.-S., He, X., Gao, J., Deng, L., Acero, A., & Heck, L. (2013) · Proceedings of the 22nd ACM international conference on Information & Knowledge Management · doi:10.1145/2505515.2505665 · Cited by 2,009 · cited 1×
“On the le, Figure 2 (a) illustrates representation-focused rankers, which independently compute an embedding for q and another for d and estimate relevance as a single similarity score between two vectors [13, 41].”
BackgroundTinyBERT: Distilling BERT for natural language understanding
Jiao, X., Yin, Y., Shang, L., Jiang, X., Chen, X., Li, L., Wang, F., & Liu, Q. (2020) · Findings of the Association for Computational Linguistics: EMNLP 2020 · doi:10.18653/v1/2020.findings-emnlp.372 · Cited by 1,786
A Deep Relevance Matching Model for Ad-hoc Retrieval
Guo, J., Fan, Y., Ai, Q., & Croft, W. B. (2016) · Proceedings of the 25th ACM International on Conference on Information and Knowledge Management · doi:10.1145/2983323.2983769 · Cited by 806 · cited 1×
“Over the past few years, the Information Retrieval (IR) community has witnessed the introduction of a host of neural ranking models, including DRMM [8], KNRM [5, 36], and Duet [21, 23].”
BackgroundNogueira, R., & Cho, K. (2019) · arXiv · doi:10.48550/arxiv.1901.04085 · Cited by 615 · cited 9×
“Similar to [25], we fine-tune all ColBERT models with learning rate 3 × 10 − 6 with a batch size 32.”
BackgroundMethodInfluentialEnd-to-End Neural Ad-hoc Ranking with Kernel Pooling
Xiong, C., Dai, Z., Callan, J., Liu, Z., & Power, R. (2017) · Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3077136.3080809 · Cited by 546 · cited 3×
“Instead of summarizing q and d into individual embeddings, these rankers model word-and phrase-level relationships across q and d and match them using a deep neural network (e.g., with CNNs/MLPs [23] or kernels [36]).”
BackgroundMethodOnline controlled experiments at large scale
Kohavi, R., Deng, A., Frasca, B., Walker, T., Xu, Y., & Pohlmann, N. (2013) · Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/2487575.2488217 · Cited by 433 · cited 1×
“This poses a challenging tradeoff since raising query response times by as little as 100ms is known to impact user experience and even measurably diminish revenue [18].”
BackgroundLearning to Match using Local and Distributed Representations of Text for Web Search
Mitra, B., Diaz, F., & Craswell, N. (2017) · Proceedings of the 26th International Conference on World Wide Web · doi:10.1145/3038912.3052579 · Cited by 426 · cited 7×
“In this work, we compare against KNRM [5, 36], Duet [21, 23], ConvKNRM [5], and fastText+ConvKNRM [11].”
BackgroundResultInfluentialConvolutional Neural Networks for Soft-Matching N-Grams in Ad-hoc Search
Dai, Z., Xiong, C., Callan, J., & Liu, Z. (2018) · Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining · doi:10.1145/3159652.3159659 · Cited by 341 · cited 9×
“To begin with, we notice the fast progress from KNRM in 2017 to the BERT-based models in 2019, manifesting itself in over 16% increase in MRR@10.”
BackgroundResultInfluentialAn Introduction to Neural Information Retrieval
Mitra, B., & Craswell, N. (2018) · Foundations and Trends® in Information Retrieval · doi:10.1561/1500000061 · Cited by 317
CEDR: Contextualized Embeddings for Document Ranking
MacAvaney, S., Yates, A., Cohan, A., & Goharian, N. (2019) · Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3331184.3331317 · Cited by 239
Anserini: Reproducible Ranking Baselines Using Lucene
Yang, P., Fang, H., & Lin, J. (2018) · Journal of Data and Information Quality · doi:10.1145/3239571 · Cited by 221
From Neural Re-Ranking to Neural Ranking: Learning a Sparse Representation for Inverted Indexing
Zamani, H., Dehghani, M., Croft, W. B., Learned-Miller, E., & Kamps, J. (2018) · Proceedings of the 27th ACM International Conference on Information and Knowledge Management · doi:10.1145/3269206.3271800 · Cited by 185
Cross-Domain Modeling of Sentence-Level Evidence for Document Retrieval
Yilmaz, Z. A., Yang, W., Zhang, H., & Lin, J. (2019) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/d19-1352 · Cited by 144 · cited 1×
“Indeed, in the span of just a few months, a number of ranking models based on BERT have achieved state-of-the-art results on various retrieval benchmarks [3, 18, 25, 39] and have been proprietarily adapted for deployment by Google1 and Bing2.”
BackgroundTREC Complex Answer Retrieval Overview.
Dietz, L., Verma, M., Radlinski, F., & Craswell, N. (2017) · Text REtrieval Conference · Cited by 84
To Index or Not to Index: Optimizing Exact Maximum Inner Product Search
Abuzaid, F., Sethi, G., Bailis, P. D., & Zaharia, M. (2019) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde.2019.00114 · Cited by 37 · cited 1×
“, [1, 15]) to retrieve the top-k results directly from a large document collection, substantially improving recall over models that only re-rank the output of term-based retrieval.”
BackgroundSelf-citationOn the Effect of Low-Frequency Terms on Neural-IR Models
Hofstätter, S., Rekabsaz, N., Eickhoff, C., & Hanbury, A. (2019) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3331184.3331344 · Cited by 35 · cited 1×
“In this work, we compare against KNRM [5, 36], Duet [21, 23], ConvKNRM [5], and fastText+ConvKNRM [11].”
ResultEfficient Interaction-based Neural Ranking with Locality Sensitive Hashing
Ji, S., Shao, J., & Yang, T. (2019) · International World Wide Web Conference (WWW) · doi:10.1145/3308558.3313576 · Cited by 26 · cited 1×
“Hofst¨atter et al. [10] and MacAvaney et al. [19] observe that BERT-based models in the literature are 100-1000 × more computationally expensive than prior models—some of which are arguably not inexpensive to begin with [14].”
Background
Research works printed but not linked: 16
- Zhuyun Dai and Jamie Callan . 2019 a. Context-Aware Sentence/Passage Term Importance Estimation For First Stage Retrieval. arXiv preprint arXiv:1910.10687 ( 2019 ). Zhuyun Dai and Jamie Callan. 2019 a. Context-Aware Sentence/Passage Term Importance Estimation For First Stage Retrieval. arXiv preprint arXiv:1910.10687 (2019). · arXiv
- Zhuyun Dai and Jamie Callan . 2019 b. Deeper Text Understanding for IR with Contextual Neural Language Modeling. arXiv preprint arXiv:1905.09217 ( 2019 ). Zhuyun Dai and Jamie Callan. 2019 b. Deeper Text Understanding for IR with Contextual Neural Language Modeling. arXiv preprint arXiv:1905.09217 (2019). · arXiv
- Jiafeng Guo , Yixing Fan , Liang Pang , Liu Yang , Qingyao Ai , Hamed Zamani , Chen Wu , W Bruce Croft , and Xueqi Cheng . 2019. A deep look into neural ranking models for information retrieval. arXiv preprint arXiv:1903.06902 ( 2019 ). Jiafeng Guo, Yixing Fan, Liang Pang, Liu Yang, Qingyao Ai, Hamed Zamani, Chen Wu, W Bruce Croft, and Xueqi Cheng. 2019. A deep look into neural ranking models for information retrieval. arXiv preprint arXiv:1903.06902 (2019). · arXiv
- Sebastian Hofst"atter and Allan Hanbury . 2019. Let's measure run time! Extending the IR replicability infrastructure to include performance aspects. arXiv preprint arXiv:1907.04614 ( 2019 ). Sebastian Hofst"atter and Allan Hanbury. 2019. Let's measure run time! Extending the IR replicability infrastructure to include performance aspects. arXiv preprint arXiv:1907.04614 (2019). · arXiv
- Sebastian Hofst"atter, Markus Zlabinger , and Allan Hanbury . 2019 b. TU Wien@ TREC Deep Learning'19--Simple Contextualization for Re-ranking. arXiv preprint arXiv:1912.01385 ( 2019 ). Sebastian Hofst"atter, Markus Zlabinger, and Allan Hanbury. 2019 b. TU Wien@ TREC Deep Learning'19--Simple Contextualization for Re-ranking. arXiv preprint arXiv:1912.01385 (2019). · arXiv
- Diederik P Kingma and Jimmy Ba . 2014 . Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014). Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014). · arXiv
- Paul Michel Omer Levy and Graham Neubig. 2019. Are Sixteen Heads Really Better than One?. In Advances in Neural Information Processing Systems. 14014--14024. Paul Michel Omer Levy and Graham Neubig. 2019. Are Sixteen Heads Really Better than One?. In Advances in Neural Information Processing Systems. 14014--14024.
- Bhaskar Mitra and Nick Craswell . 2019. An Updated Duet Model for Passage Re-ranking. arXiv preprint arXiv:1903.07666 ( 2019 ). Bhaskar Mitra and Nick Craswell. 2019. An Updated Duet Model for Passage Re-ranking. arXiv preprint arXiv:1903.07666 (2019). · arXiv
- Bhaskar Mitra , Corby Rosset , David Hawking , Nick Craswell , Fernando Diaz , and Emine Yilmaz . 2019. Incorporating query term independence assumption for efficient retrieval and ranking using deep neural networks. arXiv preprint arXiv:1907.03693 ( 2019 ). Bhaskar Mitra, Corby Rosset, David Hawking, Nick Craswell, Fernando Diaz, and Emine Yilmaz. 2019. Incorporating query term independence assumption for efficient retrieval and ranking using deep neural networks. arXiv preprint arXiv:1907.03693 (2019). · arXiv
- Rodrigo Nogueira , Wei Yang , Kyunghyun Cho , and Jimmy Lin . 2019 b. Multi-Stage Document Ranking with BERT. arXiv preprint arXiv:1910.14424 ( 2019 ). Rodrigo Nogueira, Wei Yang, Kyunghyun Cho, and Jimmy Lin. 2019 b. Multi-Stage Document Ranking with BERT. arXiv preprint arXiv:1910.14424 (2019). · arXiv
- Rodrigo Nogueira , Wei Yang , Jimmy Lin , and Kyunghyun Cho . 2019 c. Document Expansion by Query Prediction. arXiv preprint arXiv:1904.08375 ( 2019 ). Rodrigo Nogueira, Wei Yang, Jimmy Lin, and Kyunghyun Cho. 2019 c. Document Expansion by Query Prediction. arXiv preprint arXiv:1904.08375 (2019). · arXiv
- Matthew E Peters , Mark Neumann , Mohit Iyyer , Matt Gardner , Christopher Clark , Kenton Lee , and Luke Zettlemoyer . 2018. Deep contextualized word representations. arXiv preprint arXiv:1802.05365 ( 2018 ). Matthew E Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018. Deep contextualized word representations. arXiv preprint arXiv:1802.05365 (2018). · arXiv
- Yifan Qiao , Chenyan Xiong , Zhenghao Liu , and Zhiyuan Liu . 2019. Understanding the Behaviors of BERT in Ranking. arXiv preprint arXiv:1904.07531 ( 2019 ). Yifan Qiao, Chenyan Xiong, Zhenghao Liu, and Zhiyuan Liu. 2019. Understanding the Behaviors of BERT in Ranking. arXiv preprint arXiv:1904.07531 (2019). · arXiv
- Colin Raffel , Noam Shazeer , Adam Roberts , Katherine Lee , Sharan Narang , Michael Matena , Yanqi Zhou , Wei Li , and Peter J Liu . 2019. Exploring the limits of transfer learning with a unified text-to-text transformer. arXiv preprint arXiv:1910.10683 ( 2019 ). Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2019. Exploring the limits of transfer learning with a unified text-to-text transformer. arXiv preprint arXiv:1910.10683 (2019). · arXiv
- Raphael Tang , Yao Lu , Linqing Liu , Lili Mou , Olga Vechtomova , and Jimmy Lin . 2019. Distilling task-specific knowledge from BERT into simple neural networks. arXiv preprint arXiv:1903.12136 ( 2019 ). Raphael Tang, Yao Lu, Linqing Liu, Lili Mou, Olga Vechtomova, and Jimmy Lin. 2019. Distilling task-specific knowledge from BERT into simple neural networks. arXiv preprint arXiv:1903.12136 (2019). · arXiv
- Ofir Zafrir , Guy Boudoukh , Peter Izsak , and Moshe Wasserblat . 2019. Q8bert: Quantized 8bit bert. arXiv preprint arXiv:1910.06188 ( 2019 ). Ofir Zafrir, Guy Boudoukh, Peter Izsak, and Moshe Wasserblat. 2019. Q8bert: Quantized 8bit bert. arXiv preprint arXiv:1910.06188 (2019). · arXiv
Cited by
Dense passage retrieval for open-domain question answering
Karpukhin, V., Oguz, B., Min, S., Lewis, P., Wu, L., Edunov, S., Chen, D., & Yih, W.-T. (2020) · Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2020.emnlp-main.550 · Cited by 3,575
On the Opportunities and Risks of Foundation Models
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021) · arXiv · doi:10.48550/arxiv.2108.07258 · Cited by 2,676
Self-citationRetrieval-Augmented Generation for Large Language Models: A Survey
Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., Dai, Y., Sun, J., Wang, M., & Wang, H. (2023) · arXiv · doi:10.48550/arxiv.2312.10997 · Cited by 876
“The development of Advanced RAG and Modular RAG was a response to these specific shortcomings in Naive RAG.”
BackgroundRetrieval Augmentation Reduces Hallucination in Conversation
Shuster, K., Poff, S., Chen, M., Kiela, D., & Weston, J. (2021) · Findings of the Association for Computational Linguistics: EMNLP 2021 · doi:10.18653/v1/2021.findings-emnlp.320 · Cited by 575
Qu, Y., Ding, Y., Liu, J., Liu, K., Ren, R., Zhao, W. X., Dong, D., Wu, H., & Wang, H. (2021) · Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2021.naacl-main.466 · Cited by 356
Metadata
- Authors
- Khattab, O., & Zaharia, M.
- Affiliations
- Stanford University (US)
- Venue
- Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval
- Publisher
- ACM
- Published
- 2020-07-25
- Pages
- 39-48
- OpenAlex
- W3021397474
- Cited by
- 1,301
- References
- 21
- Citations by year
- 2026: 346 · 2025: 277 · 2024: 197 · 2023: 184 · 2022: 143 · 2021: 135 · 2020: 18
- FWCI
- 90.73
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) (publishedVersion)
- Funders
- NSF (CNS-1651570)
- Topics
- Physical Sciences › Computer Science › Information Systems › Information Retrieval and Search Behavior
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
- Keywords
- Colbert, BERT, document ranking, information retrieval, query-document similarity, efficient retrieval
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.