Pith. sign in

REVIEW 3 major objections 5 minor 63 references

Beyond Questions: Leveraging ColBERT for Keyphrase Search

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Keyphrase-tailored dense ranking models trained on LLM-converted keyphrase versions of MSMarco queries outperform standard ColBERT on keyphrase queries while remaining equivalent on original question queries.

desk verdict Careful empirical study with a useful frozen-encoder variant, but the keyphrase-specific gains are confounded with continued training. read the letter →

arxiv 2412.03193 v1 pith:UBK5DQYT submitted 2024-12-04 cs.IR

Jorge Gabín , Javier Parapar , Craig Macdonald This is my paper · ORCID
classification cs.IR
keywords keyphrasesearchdenseretrievalColBERTlateinteractionqueryreformulationlargelanguagemodelsMSMarcoTRECDeepLearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that standard dense retrieval models fail on keyphrase-style queries because they are trained almost entirely on question-like MSMarco queries, and that this can be fixed by training the same late-interaction architecture (ColBERT) on keyphrase-format versions of the same training triples. The authors convert MSMarco questions into keyphrases with an instruction-tuned LLM, then fine-tune either the full ColBERT model (ColBERTKP_QD) or only the query encoder (ColBERTKP_Q) on the transformed triples. On automatically generated and manually annotated keyphrase queries, both keyphrase-trained models beat standard ColBERT in end-to-end and re-ranking setups, while staying statistically equivalent on original question queries. The approach also transfers to a cross-encoder (monoT5) and improves performance on traditional title-based queries, so the claim is that query-format-specific training is a general path to better retrieval, not a quirk of one architecture. A reader should care because keyphrase search remains the norm in academic and professional search, where dense models have so far been tuned to questions.

What carries the argument

The central mechanism is Q2K, an LLM-based sequence-to-sequence transform that rewrites any query into keyphrase format, for example 'how to train a bi-encoder' becomes 'bi-encoder training'. Applying Q2K to every MSMarco training triple produces keyphrase triples <k, d+, d-> that are optimised with the standard ColBERT pairwise softmax cross-entropy, either by training both query and document encoders (ColBERTKP_QD) or by freezing the document encoder and training only the query encoder (ColBERTKP_Q). The late-interaction scoring function sums, for each query token, the maximum similarity against document tokens, so changing the query format changes which token embeddings carry the matching signal, and the paper shows the special tokens pick up the information need.

What would settle it

A direct check: take a random sample of the LLM-generated keyphrase training queries, ask human assessors whether the original relevant passage still satisfies the keyphrase query, and measure the relevance rate; if it drops markedly below the original MSMarco relevance rate, the fine-tuning signal is corrupted and the reported gains cannot be attributed to keyphrase training. Alternatively, evaluate the keyphrase-trained models on keyphrase queries produced by a different LLM or by independent annotators; if the gains over standard ColBERT disappear, the effect is an artifact of the generator rather than of keyphrase-format training.

Watch

Extended reading notes

Core claim

The paper's central claim is that training dense retrieval models on keyphrase-format queries makes them better at keyphrase search without sacrificing question-style performance. Concretely, on the TREC DL 2019 keyphrase queries, the encoder-only ColBERTKP_Q reaches MAP@1k 0.4573 and the fully trained ColBERTKP_QD 0.4505, versus 0.4303 for standard ColBERT, with parallel gains under BM25 re-ranking; a keyphrase-trained monoT5 replica shows the effect is not exclusive to late interaction. The paper reports TOST equivalence with standard ColBERT on the original TREC DL query sets, better results on manually annotated keyphrase queries and on mixed question/keyphrase scenarios, and significantly better performance on traditional title-query collections such as TREC Robust 2004 and TREC 7. It also analyses matching behaviour, finding that keyphrase training reduces reliance on lexical token overlap and shifts scoring weight into special tokens ([CLS], [Q], [SEP], [MASK]) and semantic matches.

Load-bearing premise

The load-bearing premise is that the LLM used to generate keyphrases preserves the information need of each MSMarco query, so the original relevant passages remain relevant to the converted keyphrase triples, and the paper itself notes the risk that training and test keyphrases come from the same generator.

Editorial extensions

If this is right

  • Keyphrase-tailored dense retrieval models improve ranking for keyphrase queries in both end-to-end and re-ranking setups, so deployed systems that serve professional or academic keyphrase search can expect gains from format-specific training.
  • Because the keyphrase-trained models are statistically equivalent to standard ColBERT on original question queries, the same document index can serve both query types without degrading question-based search.
  • The training approach transfers to a cross-encoder architecture (monoT5KP), indicating that the keyphrase-format training signal is useful beyond late-interaction models.
  • Gains extend to manually annotated keyphrase queries and to traditional title-format queries, so the method improves generalisability across query formats in addition to the intended keyphrase case.
  • In mixed query scenarios with half questions and half keyphrases, the keyphrase-trained models beat standard ColBERT, suggesting a single model can handle realistic heterogeneous traffic better than a question-only model.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the gains persist with a different generator, a natural next step is a query-type router that classifies incoming queries as keyphrase-style or question-style and selects the corresponding encoder, which could lift performance on real web and professional search logs without retraining document indices.
  • A testable extension is to apply the Q2K transformation to other large-scale retrieval training sets and to other languages, since the paper's evidence is currently limited to MSMarco/TREC English data.
  • The paper's observation that special tokens come to carry the information need suggests probing the [CLS] and [Q] embeddings of keyphrase-trained encoders could reveal an interpretable keyphrase-level semantics, something the paper does not analyse.
  • The synthetic 50/50 mixed-query experiment is an idealisation; real query logs are skewed and noisy, so the next evaluation should use an actual keyphrase/boolean query log to see whether the advantage survives realistic mixing ratios.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes keyphrase-tailored dense retrieval models built on ColBERT. In the absence of large keyphrase retrieval datasets, the authors use Mistral-7B-Instruct to convert MSMarco question-style queries into keyphrase-format queries, and then fine-tune ColBERT from the Wang et al. (2023) checkpoint on these transformed triples: ColBERTKP_QD (both query and document encoders trained) and ColBERTKP_Q (query encoder only, document encoder frozen). The models are evaluated on automatically generated and manually curated keyphrase versions of TREC DL 2019/2020 and MSMarco dev queries, in both end-to-end and BM25-re-ranking modes, and are also tested on title-based TREC collections and a synthetic mixed-query scenario. The central claim is Observation 1: keyphrase-tailored models outperform existing dense retrieval models on keyphrase queries while remaining roughly equivalent on original question queries.

Significance. The paper targets a real and plausible gap: dense retrievers trained predominantly on question-style queries may under-serve the keyphrase queries that remain common in academic and professional search. The contributions include a concrete data-generation pipeline, two cost/compatibility trade-off training variants, and a small manually curated keyphrase test set for TREC DL 2019. The authors release code and generated resources, make a good-faith effort to mitigate generator bias with human-annotated queries, and provide a useful analysis of matching behavior (lexical, semantic, special-token). If the central claim is substantiated, the work would be a useful empirical addition to query-formatted adaptation for dense retrieval. However, the main attribution of the reported gains to keyphrase-specific training is not yet isolated from the effect of continued fine-tuning itself, and the generalisation experiment with monoT5 has an additional confound. These issues are fixable with additional control experiments, but they are load-bearing for the headline conclusions.

major comments (3)
  1. [§5.1.4 and Tables 3, 6] The main experimental comparisons are between the original ColBERT checkpoint (Wang et al., 2023) and models fine-tuned for an additional 25k steps on MSMarco triples whose queries were converted by Mistral into keyphrases. There is no control that trains the same checkpoint for the same 25k steps on the original, unconverted MSMarco triples. Consequently, the gains reported for keyphrase queries (e.g., MAP@1k 0.4573 vs 0.4303 in Table 3; 0.3782 vs 0.3477 in Table 6) conflate two factors: the keyphrase query format and continued training itself. The manually curated test set in Table 6 addresses the same-generator evaluation bias, but it does not resolve this confound. Without the control, Observation 1 cannot be attributed specifically to keyphrase-tailored training. Please add a control experiment of the same checkpoint, the same number of steps, and the same hyperparameters on the original MSMarco triples, and report the resulting performance on keyphrase queries.
  2. [§3.2] The training procedure transforms each triple ⟨q,d+,d−⟩ into ⟨k,d+,d−⟩ by setting k = Q2K(q), but there is no verification that d+ remains relevant to the transformed keyphrase query. If the LLM conversion changes the information need or injects noise, the fine-tuning signal is corrupted, and the observed ranking improvements could reflect properties of the generator rather than of keyphrase-format relevance. The paper should quantify the faithfulness of Q2K—for example, by sampling transformed triples and having annotators (or a strong cross-encoder) check whether d+ is still relevant to k—so that the training signal can be trusted.
  3. [§5.2.3 and Table 5] The claim that keyphrase-based training generalises to monoT5 is based on a comparison between monoT5KP, trained from t5-base for 5k steps, and the monoT5 baseline, which uses a pre-trained msmarco checkpoint. The paper notes that starting from the existing monoT5 checkpoint did not help, but this still means the comparison is confounded by base model initialisation and training schedule. A cleaner experiment would apply the keyphrase transformation to the same monoT5 checkpoint and train for the same number of steps, or train both variants from t5-base on their respective data, so that any difference can be attributed to the query format rather than to the base model or number of steps.
minor comments (5)
  1. [§5.1.2] The metric abbreviation 'MMR@10' in the text should read 'MRR@10'.
  2. [§5.2.7] The sentence 'In response to the eighth research question' refers to RQ7 and should say 'seventh'.
  3. [§5.2.5] The phrase 'we visually represent each matching type using colours and different line styles in Figure 3' should refer to Figure 4, which is the figure that appears to depict the different matching types.
  4. [CRediT statement] 'Conzeptualization' is a typo and should be 'Conceptualization'.
  5. [Table 7] The 'Overlap' row in Table 7 shows values with a dash for Mistral; please clarify what the dash represents and specify the comparison (e.g., overlap of Mistral-generated keyphrases with each assessor's keyphrases).

Circularity Check

1 steps flagged · score 3.0 of 10

One acknowledged same-generator train/test loop in the automatic keyphrase evaluation; the manual keyphrase set supplies independent evidence, so the claim does not fully reduce to its inputs.

  1. other [Section 4.2 (Automatically Generating Keyphrase Queries), applied in Section 5.2.1/Table 3]
    "However, we believe that solely relying on LLM-generated data to evaluate the performance of the models is insufficient. First, there is a risk of bias in the evaluation process, as the similarity between the training and testing queries—both generated by the same model—could affect the results."

    The automatic test keyphrases in Table 3 are generated with the same Mistral Q2K prompt used to build the training triples (Section 3.2), so the reported gains on keyphrase queries partly measure the models' fit to Mistral's keyphrase style rather than a general keyphrase-query capability. The paper's own words admit this risk. It is not an equation-level reduction or a label leak, and Section 5.2.4's human-annotated keyphrases provide an independent check, so the circularity is partial and acknowledged rather than a collapse of the central claim.

full rationale

This is an experimental paper, not a derivation, so no equation equates a predicted quantity to a fitted input. The only load-bearing evaluation concern is the same-generator overlap: training uses Mistral Q2K(MSMarco queries) and the automatic test uses Mistral Q2K(TREC queries), making the automatic keyphrase benchmark partly in-distribution for the proposed models. The paper explicitly flags this and counters it with a manually curated keyphrase test set (Section 4.3, Table 6), where keyphrase-tailored models still beat ColBERT on all metrics; this independent evidence prevents the central claim from reducing to the generator's style. The skeptic's missing-control point (no 25k-step fine-tune on original MSMarco triples) is a causal-attribution limitation, not a circularity: Table 4 shows equivalence on original queries, so continued training alone does not explain the keyphrase-query gains. No self-citation chain is load-bearing, and all external benchmarks use TREC qrels.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claims rely on four domain assumptions about query transformation and label transfer, plus the hand-designed Mistral prompt. There are no invented physical entities and no fitted model parameters beyond standard fine-tuning hyperparameters.

free parameters (2)
  • Mistral Q2K prompt template = Provided in Appendix B (3 in-context examples, output format <Keyphrases>...)
    Hand-designed prompt that determines the distribution of all training and test keyphrase queries; no ablation over prompt variants is reported.
  • Manual keyphrase evaluation policy = First keyphrase per assessor only
    Used for RQ4 and mixed-query experiments; discards multiple keyphrases from Assessor 1 (avg 2.91 per query), which may under-represent how humans formulate keyphrase queries.
assumptions (4)
  • domain assumption MSMarco relevance labels transfer to Q2K(MSMarco) queries
    Training triples use the original relevant passage with the transformed keyphrase query; no validation that d+ remains relevant to k (Section 3.2).
  • domain assumption LLM-generated keyphrases approximate real keyphrase queries
    The paper's main evaluation relies on Mistral-generated test keyphrases. It checks overlap with manual annotations (~75% term overlap, Table 7) but does not guarantee real-user behavior.
  • domain assumption Keyphrase queries are a distinct and important query class
    Motivated by citations (Jacsó 2015, Russell-Rose 2018) and the MSMarco question-rate analysis in Table 2; not directly measured in this paper.
  • domain assumption Short title queries (TREC 7/8/Robust) are representative of keyphrase queries
    Used for RQ7 generalisation; title queries are short but not necessarily the same distribution as user keyphrase queries.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Questions: Leveraging ColBERT for Keyphrase Search." pith.science (2026). https://pith.science/paper/UBK5DQYT

@misc{pith2026241203193,
  author       = {Pith},
  title        = {Pith review of: Beyond Questions: Leveraging ColBERT for Keyphrase Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UBK5DQYT}},
  note         = {Machine review of arXiv:2412.03193}
}
read the original abstract

While question-like queries are gaining popularity and search engines' users increasingly adopt them, keyphrase search has traditionally been the cornerstone of web search. This query type is also prevalent in specialised search tasks such as academic or professional search, where experts rely on keyphrases to articulate their information needs. However, current dense retrieval models often fail with keyphrase-like queries, primarily because they are mostly trained on question-like ones. This paper introduces a novel model that employs the ColBERT architecture to enhance document ranking for keyphrase queries. For that, given the lack of large keyphrase-based retrieval datasets, we first explore how Large Language Models can convert question-like queries into keyphrase format. Then, using those keyphrases, we train a keyphrase-based ColBERT ranker (ColBERTKP_QD) to improve the performance when working with keyphrase queries. Furthermore, to reduce the training costs associated with training the full ColBERT model, we investigate the feasibility of training only a keyphrase query encoder while keeping the document encoder weights static (ColBERTKP_Q). We assess our proposals' ranking performance using both automatically generated and manually annotated keyphrases. Our results reveal the potential of the late interaction architecture when working under the keyphrase search scenario.

Figures

Figures reproduced from arXiv: 2412.03193 by the authors.

Figure 1
Figure 1. Training process of the ColBERTKP𝑄 model using the transformed MSMarco training triples. There are differences in matching behaviour between question-like queries and keyphrase ones that justify the need for training specific models. Question-like queries usually contain interrogative adverb phrases (e.g., "when", "where", "how", etc.), which guide the model towards the information need and might match with answer w… view at source ↗
Figure 2
Figure 2. ColBERT and ColBERTKP𝑄 interaction for query 962179 (in both original and keyphrase format) and passage 2329699 (shortened). Darker shading in the interaction matrix denotes higher similarity, while the × symbol highlights the document embedding (row) with the highest similarity for each query embedding (column). The histogram above illustrates each query embedding’s contribution to the documents’s final score, with… view at source ↗
Figure 3
Figure 3. Performance across different types of queries according to the query type taxonomy presented by Bolotova et al. (2022). Original and Mistral keyphrases bars show the delta between using ColBERTKP𝑄 and ColBERT as the retrieval model. ColBERTKP𝑄 and ColBERT bars show the delta between using keyphrases and original queries. Observation 4 The superior performance of keyphrase-based models using human-generated keyphrase… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Representation of each matching type. Our second experiment examines how training ColBERT on keyphrases affects its matching behaviour. Following the methodology of Wang et al. (2023), we analyse the three matching strategies between query and document. Lexical matchin…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 29 canonical work pages

  1. [1]

    Dalton , author S

    author J. Dalton , author S. Fischer , author P. Owoicho , author F. Radlinski , author F. Rossetto , author J. R. Trippas , author H. Zamani , title Conversational information seeking: Theory and application , in: booktitle Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval , SIGIR '22, publish...

    arXiv 2022

  2. [2]

    GUO , author C

    author K. GUO , author C. Defretiere , author D. Diefenbach , author C. Gravier , author A. Gourru , title Qanswer: Towards question answering search over websites , in: booktitle Companion Proceedings of the Web Conference 2022 , WWW '22, publisher Association for Computing Machinery , address New York, NY, USA , year 2022 , p. pages 252–255 . https://do...

    arXiv 2022

  3. [3]

    Khurana , author A

    author D. Khurana , author A. Koli , author K. Khatter , author S. Singh , title Natural language processing: state of the art, current trends and challenges , journal Multimedia Tools and Applications volume 82 ( year 2023 ) pages 3713--3744

    work page 2023

  4. [4]

    Jacs \' o , title Academic search engines: A quantitative outlook , journal Online Inf

    author P. Jacs \' o , title Academic search engines: A quantitative outlook , journal Online Inf. Rev. volume 39 ( year 2015 ) pages 435--436

    work page 2015

  5. [5]

    Li , author B

    author X. Li , author B. J. Schijvenaars , author M. de Rijke , title Investigating queries and search failures in academic search , journal Information Processing & Management volume 53 ( year 2017 ) pages 666--683

    work page 2017

  6. [6]

    Russell - Rose , author J

    author T. Russell - Rose , author J. Chamberlain , author L. Azzopardi , title Information retrieval in the workplace: A comparison of professional search practices , journal Information Processing & Management volume 54 ( year 2018 ) pages 1042--1057

    work page 2018

  7. [7]

    Lahiri , author D

    author A. Lahiri , author D. K. Sanyal , author I. Mukherjee , title A keyphrase-centric search engine for scientific papers , in: booktitle Proceedings of the 15th Annual Meeting of the Forum for Information Retrieval Evaluation , FIRE '23, publisher Association for Computing Machinery , address New York, NY, USA , year 2024 , p. pages 125–128 . https://...

    arXiv 2024

  8. [8]

    Nguyen , author M

    author T. Nguyen , author M. Rosenberg , author X. Song , author J. Gao , author S. Tiwary , author R. Majumder , author L. Deng , title MS MARCO: A human generated machine reading comprehension dataset , in: editor T. R. Besold , editor A. Bordes , editor A. S. d'Avila Garcez , editor G. Wayne (Eds.), booktitle Proceedings of the Workshop on Cognitive Co...

    work page 2016

Show all 63 references
  1. [9]

    Vaswani , author N

    author A. Vaswani , author N. Shazeer , author N. Parmar , author J. Uszkoreit , author L. Jones , author A. N. Gomez , author L. Kaiser , author I. Polosukhin , title Attention is all you need , in: editor I. Guyon , editor U. von Luxburg , editor S. Bengio , editor H. M. Wal...

    2017

  2. [10]

    Devlin , author M

    author J. Devlin , author M. Chang , author K. Lee , author K. Toutanova , title BERT: pre-training of deep bidirectional transformers for language understanding , in: editor J. Burstein , editor C. Doran , editor T. Solorio (Eds.), booktitle Proceedings of the 2019 Conference...

    2019

  3. [11]

    Karpukhin , author B

    author V. Karpukhin , author B. Oguz , author S. Min , author P. S. H. Lewis , author L. Wu , author S. Edunov , author D. Chen , author W. Yih , title Dense passage retrieval for open-domain question answering , in: editor B. Webber , editor T. Cohn , editor Y. He , editor Y....

    2020

  4. [12]

    Xiong , author C

    author L. Xiong , author C. Xiong , author Y. Li , author K. Tang , author J. Liu , author P. N. Bennett , author J. Ahmed , author A. Overwijk , title Approximate nearest neighbor negative contrastive learning for dense text retrieval , in: booktitle 9th International Confere...

    2021

  5. [13]

    Khattab , author M

    author O. Khattab , author M. Zaharia , title Colbert: Efficient and effective passage search via contextualized late interaction over BERT , in: editor J. X. Huang , editor Y. Chang , editor X. Cheng , editor J. Kamps , editor V. Murdock , editor J. Wen , editor Y. Liu (Eds.)...

    2020

  6. [15]

    Ni , author C

    author J. Ni , author C. Qu , author J. Lu , author Z. Dai , author G. H. \' A brego , author J. Ma , author V. Y. Zhao , author Y. Luan , author K. B. Hall , author M. Chang , author Y. Yang , title Large dual encoders are generalizable retrievers , in: editor Y. Goldberg , e...

    2022

  7. [16]

    Lee , author Z

    author J. Lee , author Z. Dai , author S. M. K. Duddu , author T. Lei , author I. Naim , author M. Chang , author V. Zhao , title Rethinking the role of token retrieval in multi-vector retrieval , in: editor A. Oh , editor T. Naumann , editor A. Globerson , editor K. Saenko , ...

    2023

  8. [17]

    Douze , author A

    author M. Douze , author A. Guzhva , author C. Deng , author J. Johnson , author G. Szilvasy , author P. Mazar \' e , author M. Lomeli , author L. Hosseini , author H. J \' e gou , title The faiss library , journal CoRR volume abs/2401.08281 ( year 2024 )

    2024 arXiv

  9. [18]

    Johnson , author M

    author J. Johnson , author M. Douze , author H. J \' e gou , title Billion-scale similarity search with gpus , journal IEEE Trans. Big Data volume 7 ( year 2021 ) pages 535--547

    2021

  10. [19]

    Silverstein , author M

    author C. Silverstein , author M. Henzinger , author H. Marais , author M. Moricz , title Analysis of a very large web search engine query log , volume volume 33 , year 1999 , pp. pages 6--12 . https://doi.org/10.1145/331403.331405. :10.1145/331403.331405

    1999

  11. [20]

    author J. H. Reimer , author S. Schmidt , author M. Fr \" o be , author L. Gienapp , author H. Scells , author B. Stein , author M. Hagen , author M. Potthast , title The archive query log: Mining millions of search result pages of hundreds of search engines from 25 years of w...

    2023

  12. [21]

    Taghavi , author A

    author M. Taghavi , author A. Patel , author N. Schmidt , author C. Wills , author Y. Tew , title An analysis of web proxy logs with query distribution pattern approach for search engines , journal Computer Standards & Interfaces volume 34 ( year 2012 ) pages 162--170

    2012

  13. [22]

    Bolotova , author V

    author V. Bolotova , author V. Blinov , author F. Scholer , author W. B. Croft , author M. Sanderson , title A non-factoid question-answering taxonomy , in: editor E. Amig \' o , editor P. Castells , editor J. Gonzalo , editor B. Carterette , editor J. S. Culpepper , editor G....

    2022

  14. [23]

    author R. F. Nogueira , author Z. Jiang , author R. Pradeep , author J. Lin , title Document ranking with a pretrained sequence-to-sequence model , in: editor T. Cohn , editor Y. He , editor Y. Liu (Eds.), booktitle Findings of the Association for Computational Linguistics: EM...

    2020 doi

  15. [24]

    Raffel , author N

    author C. Raffel , author N. Shazeer , author A. Roberts , author K. Lee , author S. Narang , author M. Matena , author Y. Zhou , author W. Li , author P. J. Liu , title Exploring the limits of transfer learning with a unified text-to-text transformer , journal J. Mach. Learn....

    2020

  16. [25]

    Pradeep , author R

    author R. Pradeep , author R. F. Nogueira , author J. Lin , title The expando-mono-duo design pattern for text ranking with pretrained sequence-to-sequence models , journal CoRR volume abs/2101.05667 ( year 2021 )

    2021 arXiv

  17. [26]

    Macdonald , author N

    author C. Macdonald , author N. Tonellotto , author I. Ounis , title On single and multiple representations in dense passage retrieval , in: editor V. W. Anelli , editor T. D. Noia , editor N. Ferro , editor F. Narducci (Eds.), booktitle Proceedings of the 11th Italian Informa...

    2021

  18. [27]

    Lin , author J

    author S. Lin , author J. Yang , author J. Lin , title In-batch negatives for knowledge distillation with tightly-coupled teachers for dense retrieval , in: editor A. Rogers , editor I. Calixto , editor I. Vulic , editor N. Saphra , editor N. Kassner , editor O. Camburu , edit...

    2021

  19. [28]

    author Y. Kim , title Applications and future of dense retrieval in industry , in: booktitle Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval , SIGIR '22, publisher Association for Computing Machinery , address New...

    2022

  20. [29]

    Clavi \'e , title Jacolbertv2

    author B. Clavi \'e , title Jacolbertv2. 5: Optimising multi-vector retrievers to create state-of-the-art japanese retrievers with constrained resources , journal arXiv preprint arXiv:2407.20750 ( year 2024 )

    2024 arXiv

  21. [30]

    Lee , author A

    author S. Lee , author A. Shakir , author D. Koenig , author J. Lipp , title Colbertus maximus - introducing mxbai-colbert-large-v1 , year 2024 . https://www.mixedbread.ai/blog/mxbai-colbert-large-v1, note accessed: 2024-10-16

    2024

  22. [31]

    Jha , author B

    author R. Jha , author B. Wang , author M. G \"u nther , author S. Sturua , author M. K. Akram , author H. Xiao , title Jina-colbert-v2: A general-purpose multilingual late interaction retriever , journal arXiv preprint arXiv:2408.16672 ( year 2024 )

    2024 arXiv

  23. [32]

    note Accessed: 2024-10-16

    author Intel , title Colbert-nq , howpublished https://huggingface.co/Intel/ColBERT-NQ , year 2024 . note Accessed: 2024-10-16

    2024

  24. [33]

    a tter , author S. Althammer , author M. Schr \

    author S. Hofst \" a tter , author S. Althammer , author M. Schr \" o der , author M. Sertkan , author A. Hanbury , title Improving efficient neural ranking models with cross-architecture knowledge distillation , journal CoRR volume abs/2010.02666 ( year 2020 )

    2010 arXiv

  25. [35]

    Lin , author J

    author S. Lin , author J. Yang , author J. Lin , title Distilling dense representations for ranking using tightly-coupled teachers , journal CoRR volume abs/2010.11386 ( year 2020 )

    2010 arXiv

  26. [36]

    Wang , author S

    author X. Wang , author S. MacAvaney , author C. Macdonald , author I. Ounis , title An inspection of the reproducibility and replicability of tct-colbert , in: editor E. Amig \' o , editor P. Castells , editor J. Gonzalo , editor B. Carterette , editor J. S. Culpepper , edito...

    2022

  27. [37]

    Zhan , author J

    author J. Zhan , author J. Mao , author Y. Liu , author J. Guo , author M. Zhang , author S. Ma , title Optimizing dense retrieval model training with hard negatives , in: editor F. Diaz , editor C. Shah , editor T. Suel , editor P. Castells , editor R. Jones , editor T. Sakai...

    2021

  28. [38]

    Lassance , author M

    author C. Lassance , author M. Maachou , author J. Park , author S. Clinchant , title Learned token pruning in contextualized late interaction over BERT (colbert) , in: editor E. Amig \' o , editor P. Castells , editor J. Gonzalo , editor B. Carterette , editor J. S. Culpepper...

    2022

  29. [39]

    Santhanam , author O

    author K. Santhanam , author O. Khattab , author J. Saad - Falcon , author C. Potts , author M. Zaharia , title Colbertv2: Effective and efficient retrieval via lightweight late interaction , in: editor M. Carpuat , editor M. de Marneffe , editor I. V. M. Ru \' z (Eds.), bookt...

    2022

  30. [40]

    Kulkarni , author D

    author M. Kulkarni , author D. Mahata , author R. Arora , author R. Bhowmik , title Learning rich representation of keyphrases from text , in: editor M. Carpuat , editor M. de Marneffe , editor I. V. M. Ru \' z (Eds.), booktitle Findings of the Association for Computational Li...

    2022 doi

  31. [41]

    Gab \' n , author M

    author J. Gab \' n , author M. E. Ares , author J. Parapar , title Exploring models for automatic keyword labelling of scientific documents , in: editor L. Tamine , editor E. Amig \' o , editor J. Mothe (Eds.), booktitle Proceedings of the 2nd Joint Conference of the Informati...

    2022

  32. [42]

    Song , author H

    author M. Song , author H. Jiang , author S. Shi , author S. Yao , author S. Lu , author Y. Feng , author H. Liu , author L. Jing , title Is chatgpt A good keyphrase generator? A preliminary study , journal CoRR volume abs/2303.13001 ( year 2023 )

    2023 arXiv

  33. [43]

    Mart \' nez - Cruz , author A

    author R. Mart \' nez - Cruz , author A. J. L \' o pez - L \' o pez , author J. Portela , title Chatgpt vs state-of-the-art models: A benchmarking study in keyphrase generation task , journal CoRR volume abs/2304.14177 ( year 2023 )

    2023 arXiv

  34. [44]

    Bennani - Smires , author C

    author K. Bennani - Smires , author C. Musat , author A. Hossmann , author M. Baeriswyl , author M. Jaggi , title Simple unsupervised keyphrase extraction using sentence embeddings , in: editor A. Korhonen , editor I. Titov (Eds.), booktitle Proceedings of the 22nd Conference ...

    2018 doi

  35. [45]

    author I. H. Witten , author G. W. Paynter , author E. Frank , author C. Gutwin , author C. G. Nevill-Manning , title Kea: Practical automatic keyphrase extraction , in: booktitle Proceedings of the fourth ACM conference on Digital libraries , year 1999 , pp. pages 254--255

    1999

  36. [46]

    Wang , author C

    author X. Wang , author C. MacDonald , author N. Tonellotto , author I. Ounis , title Colbert-prf: Semantic pseudo-relevance feedback for dense passage and document retrieval , journal ACM Trans. Web volume 17 ( year 2023 a ) pages 3:1--3:39

    2023

  37. [47]

    Wang , author S

    author X. Wang , author S. MacAvaney , author C. Macdonald , author I. Ounis , title Effective contrastive weighting for dense query expansion , in: editor A. Rogers , editor J. L. Boyd - Graber , editor N. Okazaki (Eds.), booktitle Proceedings of the 61st Annual Meeting of th...

    2023

  38. [48]

    Jagerman , author H

    author R. Jagerman , author H. Zhuang , author Z. Qin , author X. Wang , author M. Bendersky , title Query expansion by prompting large language models , journal CoRR volume abs/2305.03653 ( year 2023 )

    2023 arXiv

  39. [49]

    Wang , author N

    author L. Wang , author N. Yang , author F. Wei , title Query2doc: Query expansion with large language models , in: editor H. Bouamor , editor J. Pino , editor K. Bali (Eds.), booktitle Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMN...

    2023

  40. [50]

    Alaofi , author L

    author M. Alaofi , author L. Gallagher , author M. Sanderson , author F. Scholer , author P. Thomas , title Can generative llms create query variants for test collections? an exploratory study , in: editor H. Chen , editor W. E. Duh , editor H. Huang , editor M. P. Kato , edit...

    2023

  41. [51]

    Nogueira , author J

    author R. Nogueira , author J. Lin , author A. Epistemic , title From doc2query to doctttttquery , journal Online preprint volume 6 ( year 2019 )

    2019

  42. [52]

    author B. B. Cambazoglu , author L. Tavakoli , author F. Scholer , author M. Sanderson , author W. B. Croft , title An intent taxonomy for questions asked in web search , in: editor F. Scholer , editor P. Thomas , editor D. Elsweiler , editor H. Joho , editor N. Kando , editor...

    2021

  43. [53]

    author A. Q. Jiang , author A. Sablayrolles , author A. Mensch , author C. Bamford , author D. S. Chaplot , author D. de Las Casas , author F. Bressand , author G. Lengyel , author G. Lample , author L. Saulnier , author L. R. Lavaud , author M. Lachaux , author P. Stock , aut...

    arXiv

  44. [54]

    author E. M. Voorhees , title The trec robust retrieval track , journal SIGIR Forum volume 39 ( year 2005 ) pages 11–20

    2005

  45. [55]

    Hawking , author N

    author D. Hawking , author N. Craswell , author P. B. Thistlewaite , title Overview of TREC-7 very large collection track , in: editor E. M. Voorhees , editor D. K. Harman (Eds.), booktitle Proceedings of The Seventh Text REtrieval Conference, TREC 1998, Gaithersburg, Maryland...

    1998

  46. [56]

    Hawking , author E

    author D. Hawking , author E. M. Voorhees , author N. Craswell , author P. Bailey , title Overview of the TREC-8 web track , in: editor E. M. Voorhees , editor D. K. Harman (Eds.), booktitle Proceedings of The Eighth Text REtrieval Conference, TREC 1999, Gaithersburg, Maryland...

    1999

  47. [57]

    Craswell , author B

    author N. Craswell , author B. Mitra , author E. Yilmaz , author D. Campos , author E. M. Voorhees , title Overview of the TREC 2019 deep learning track , journal CoRR volume abs/2003.07820 ( year 2020 a )

    2019 arXiv

  48. [58]

    Craswell , author B

    author N. Craswell , author B. Mitra , author E. Yilmaz , author D. Campos , title Overview of the TREC 2020 deep learning track , in: editor E. M. Voorhees , editor A. Ellis (Eds.), booktitle Proceedings of the Twenty-Ninth Text REtrieval Conference, TREC 2020, Virtual Event ...

    2020

  49. [59]

    Macdonald , author N

    author C. Macdonald , author N. Tonellotto , title Declarative experimentation in information retrieval using pyterrier , in: editor K. Balog , editor V. Setty , editor C. Lioma , editor Y. Liu , editor M. Zhang , editor K. Berberich (Eds.), booktitle ICTIR '20: The 2020 ACM S...

    2020

  50. [60]

    author D. J. Schuirmann , title A comparison of the two one-sided tests procedure and the power approach for assessing the equivalence of average bioavailability , journal Journal of pharmacokinetics and biopharmaceutics volume 15 ( year 1987 ) pages 657--680

    1987

  51. [61]

    Dettmers , author L

    author T. Dettmers , author L. Zettlemoyer , title The case for 4-bit precision: k-bit inference scaling laws , in: editor A. Krause , editor E. Brunskill , editor K. Cho , editor B. Engelhardt , editor S. Sabato , editor J. Scarlett (Eds.), booktitle International Conference ...

    2023

  52. [62]

    Pradeep , author H

    author R. Pradeep , author H. Chen , author L. Gu , author M. S. Tamber , author J. Lin , title Pygaggle: A gaggle of resources for open-domain question answering , in: editor J. Kamps , editor L. Goeuriot , editor F. Crestani , editor M. Maistro , editor H. Joho , editor B. D...

    2023

  53. [63]

    Formal , author B

    author T. Formal , author B. Piwowarski , author S. Clinchant , title Match your words! A study of lexical matching in neural information retrieval , in: editor M. Hagen , editor S. Verberne , editor C. Macdonald , editor C. Seifert , editor K. Balog , editor K. N rv g , edito...

    2022

  54. [64]

    Formal , author B

    author T. Formal , author B. Piwowarski , author S. Clinchant , title A white box analysis of colbert , in: booktitle Advances in Information Retrieval: 43rd European Conference on IR Research, ECIR 2021, Virtual Event, March 28--April 1, 2021, Proceedings, Part II 43 , organi...

    2021

  55. [65]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.