Interpretability Analysis of Domain Adapted Dense Retrievers

Goksenin Yuksel; Jaap Kamps

REVIEW 3 major objections 5 minor 22 references

Interpretability Analysis of Domain Adapted Dense Retrievers

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read By applying Integrated Gradients to dense retrievers, this paper shows that unsupervised domain adaptation shifts token-level attribution toward domain-specific vocabulary, and argues that Integrated Gradients gives a viable window into…

desk verdict A plausible, honest exploratory study applying Integrated Gradients to dense retrievers, but the main generalization rests on frequency-confounded word clouds and two anecdotal queries. read the letter →

arxiv 2501.14459 v1 pith:YAMWBQ2I submitted 2025-01-24 cs.IR cs.AI

Goksenin Yuksel , Jaap Kamps This is my paper · ORCID

classification cs.IRcs.AI

keywords integratedgradientsdenseretrievaldomainadaptationinterpretabilityinputattributionunsupervisedneuralinformation

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

Dense retrievers map queries and documents into vectors, but what they actually attend to is hard to see. This paper adapts the Integrated Gradients attribution method to dense retrieval, with a new baseline that attributes importance separately to query and document tokens, and applies it to models before and after unsupervised domain adaptation on a financial and a biomedical dataset. The central finding is that the adapted models place more positive attribution on in-domain terminology such as "hedge," "gold," "corona," and "disease" than the unadapted base models do. The paper also argues that Integrated Gradients is a viable tool for explaining dense retrievers, including both single-pair (instance) explanations and explanations of an entire ranking. If correct, this gives IR practitioners a way to inspect what domain adaptation changes inside the model.

What carries the argument

The central object is a modified Integrated Gradients baseline for dense retrievers. Integrated Gradients attributes a model's output to input tokens by integrating gradients along a straight path from a baseline input to the actual input; the paper's baseline uses padding tokens, replacing query tokens while keeping the document fixed to get query attributions, and replacing document tokens while keeping the query fixed to get document attributions. This yields instance-based explanations for a single query-document pair. For ranking-based explanations, the method sums token attributions over the top 25 retrieved documents for a query, so tokens that recur in highly ranked documents and contribute positively receive large aggregated scores. This aggregation turns the attribution method into a tool for comparing how base and domain-adapted models weight vocabulary across an entire ranked list.

What would settle it

Compute token attributions for a larger, randomly drawn set of queries (say 100 per dataset) for both the adapted and the base model, define a fixed list of in-domain terms per dataset, and test whether the mean positive attribution for those terms is statistically significantly higher in the adapted model; if the difference is absent or within noise, the paper's qualitative pattern would not survive in aggregate.

Watch

Extended reading notes

Core claim

The paper's central claim is that Integrated Gradients can be used to produce token-level explanations for dense retrievers, and that applying this method reveals a consistent behavioral shift after domain adaptation: the adapted model concentrates its positive attributions on vocabulary specific to the target domain, while the unadapted model attends to more generic terms. To obtain query and document attributions, the paper replaces query tokens with padding tokens while leaving the document unchanged, and vice versa, running integrated gradients from those baselines. It then aggregates token attributions over the top 25 retrieved documents to produce ranking-level word clouds. On TREC-COVID and FIQA, the attributed tokens shift from generic words toward terms like "corona," "disease," "hedge," and "gold," and for TREC-COVID the adapted model assigns positive attribution to document titles where the base model assigns negative attribution. The paper interprets this as evidence that unsupervised domain adaptation teaches the retriever to rely on domain vocabulary and document structure rather than generic overlap.

Load-bearing premise

The central observation rests on a small set of example queries, which the paper itself calls a deep qualitative analysis of a small sample of queries, with one query-document pair per model per dataset and attributions summed over the top 25 retrieved documents, so the claim that domain adaptation shifts attribution to in-domain terms depends on those examples being representative.

Editorial extensions

If this is right

If domain-adapted retrievers focus on in-domain terms, then explanations of retrieved results can be read as evidence about what vocabulary the model learned, not just which documents match.
The proposed baseline gives a reusable way to attribute dense retrieval scores to query and document tokens, applicable to any dot-product bi-encoder.
Ranking-level attributions let practitioners inspect not only a single match but the set of top documents, revealing terms that drive an entire ranking.
For TREC-COVID specifically, the title attribution result suggests that unsupervised adaptation can make a model rely more heavily on document structure (titles), a behavior with consequences for how documents should be formatted for retrieval.

Reading between the lines

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

A direct next step would be to convert the word-cloud comparisons into a quantitative attribution-difference score over a fixed domain vocabulary list; the paper notes that no standard quantitative evaluation of attributions exists, so such a metric would be a natural extension.
If the title-attribution result generalizes, then unsupervised adaptation on title-concatenated documents may teach retrievers to lean on titles; one way to test this is to remove titles at inference time and compare effectiveness drops for adapted and base models.
The same baseline-based attribution procedure could be applied to other unsupervised adaptation objectives, such as query generation or knowledge distillation, to see whether the in-domain vocabulary focus is specific to pseudo-labeling or common to all adaptation methods.
Because the paper's ranking attributions are averages over top documents, a fully global explanation could be built by averaging over many queries and documents; the paper flags global attributions as out of reach for Integrated Gradients, so this would be a different tool rather than a trivial extension.

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.

Desk Editor's Note

read the letter

The paper applies Integrated Gradients to dense bi-encoder retrieval using a separate [PAD] baseline for query and document tokens. That is a genuine extension of Zhan et al.'s cross-encoder work, and it is the first application to dense retrievers as far as I can tell. The method section is clear, and the retrieval numbers in Table 1 reproduce the known GPL gains, which is a good sanity check. The title attribution analysis in Figure 5 is a nice quantitative addition, even if small in scope.

The paper is also honest about its limitations. The appendix concedes a small sample and the absence of a standard quantitative evaluation of attributions. That is real intellectual honesty, and it makes the work easier to trust as an exploratory proof of concept.

The soft spots are real, though. The ranking word clouds sum attributions over the top 25 retrieved documents. That conflates per-occurrence attribution with term frequency: a word like "gold" can appear large simply because it shows up in many of the top 25 documents, not because the model gives it high attribution per occurrence. The stress-test note is right that the central claim about domain-adapted models focusing more on in-domain terminology needs a frequency-controlled test. The paper does not provide one. Also, the "random" queries and documents are not described with seeds or selection criteria, so I cannot verify that the base and adapted models are being compared on identical inputs. The instance-based figures are anecdotal, as the authors admit.

None of this sinks the paper. The methodological claim that IG is viable for dense retrieval explanation holds up as a proof of concept. The empirical claim about what domain adaptation does is plausible but not established. I would not cite it as evidence for that claim, but I would cite it as prior work using this baseline.

This is a workshop-scale exploratory study. A serious referee should engage with it, and the right outcome would be a revision that adds a frequency-controlled analysis and releases code and data. For a full conference I would be skeptical; for a workshop or short-paper track I would lean accept.

Referee Report

3 major / 5 minor

Summary. The paper proposes an Integrated Gradients (IG) based attribution method for dense retrievers, introducing a PAD-token baseline that yields query-side and document-side token attributions. It applies the method to compare a MS MARCO fine-tuned DistilBERT dense retriever with its GPL domain-adapted variants on TREC-COVID and FIQA, reporting both instance-level and ranking-level (top-25 aggregated) attribution visualizations. The authors reproduce the known GPL effectiveness gains and qualitatively observe that domain-adapted models assign more positive attribution to in-domain terminology such as "corona," "hedge," and "gold," and (for TREC-COVID) to document titles. They conclude that IG is a viable interpretability tool for dense retrievers and that domain adaptation shifts token-level attribution toward domain-specific vocabulary.

Significance. If the central observations were quantitatively established, the paper would provide a useful starting point for understanding how unsupervised domain adaptation changes dense retriever behavior at the token level, and the proposed PAD baseline is a reasonable, axiomatically grounded way to obtain query and document attributions. The paper uses public models and datasets, reproduces known retrieval-effectiveness gains in Table 1, and is transparent about its limitations. The strengths are the simplicity of the attribution setup and the explicit admission of the exploratory nature of the analysis. However, the significance is currently limited because the main empirical claim rests on a small number of qualitative examples and an aggregation procedure that has not been shown to isolate token importance from token frequency.

major comments (3)

[Section 3, Ranking Analysis; Figures 1-4] The ranking-based attribution word clouds are produced by summing token attributions over the top 25 retrieved documents. This conflates per-occurrence attribution with token frequency: a token such as "gold" can appear larger because it occurs in many of the top documents rather than because the model assigns it higher importance per occurrence. The paper's central claim that domain-adapted models "focus more on in-domain terminology" is therefore not established by the visualized word clouds. A frequency-controlled test is needed, for example average attribution per token occurrence, attribution conditioned on document frequency, or a comparison between the base and adapted models on matched document sets.
[Section 4.1 and 4.2; Figures 1-4] The qualitative comparison between base and domain-adapted models is based on one "random" query-document pair per dataset per model, with no seed, sampling protocol, or query/document identifiers reported. The captions do not state whether Figures 1 and 2 (and Figures 3 and 4) use the same query and document for the two models, although the text implies such a comparison. Without this information, the observed attribution differences cannot be verified or attributed to domain adaptation rather than to input variation and selection bias. A larger, documented sample of queries with quantitative aggregation (e.g., mean attribution for domain terms over many queries, with confidence intervals) is required to support the general claim.
[A Limitations; Section 5] The paper concedes that there is "no standard quantitative way of evaluating attributions" and that the analysis is "a deep qualitative analysis of a small sample of queries." This is honest, but the central claim about attribution shifts after domain adaptation can still be quantified without solving the general faithfulness problem: for instance, by computing aggregate attribution statistics over a random sample of test queries, comparing per-token or per-term attribution between base and adapted models, and reporting significance. Figure 5 is a step in this direction but reports no sample size, variance, or statistical test. Adding such a quantitative layer is essential for the paper's main conclusion.

minor comments (5)

[Table 1] The column header "GLP" should be "GPL"; the same typo should be corrected throughout if it appears elsewhere.
[Figure captions 1-4] The captions say the query and document are "random" but do not state the random seed, the sampling procedure, or whether the same query-document pair is used for the base and domain-adapted models; please add this information and ideally make the exact inputs available.
[Figure 5] The title-attribution plot would benefit from error bars, the number of query-document pairs used, and a statistical comparison (e.g., a paired test) between the base and adapted models.
[Section 3, Baseline] The sentence "we evaluate provided the models" is ungrammatical and should be revised, and the pooling/sequence-length settings should be stated more precisely with respect to the original GPL configuration.
[General] The paper uses placeholder ACM template text (e.g., the header "Trovato et al." and conference name placeholders) that should be removed or replaced before submission.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the IG attribution analysis is an external post-hoc comparison on fixed models, with no fitted parameter renamed as a prediction and no load-bearing self-citation.

full rationale

The paper contains no fitted parameter that is later called a prediction, no derivation that defines its target in terms of its inputs, and no load-bearing self-citation. The integrated gradients attributions are computed post hoc on fixed, externally provided model checkpoints (MS MARCO fine-tuned DistilBERT and GPL-adapted versions), using a clearly stated PAD-replacement baseline and the standard integrated gradients formula. The comparison between base and domain-adapted models is an external empirical comparison, not an analytic identity. The title-attribution aggregate in Figure 5 sums the same attribution scores and is not used to define or fit the models, so it does not reduce the conclusion to its input. The limitations section explicitly concedes that the evaluation is a deep qualitative analysis of a small sample of queries and that there is no standard quantitative way of evaluating attributions; this is an honest generalizability limitation, not circularity. The ranking word-cloud aggregation sums attributions over the top 25 retrieved documents, which can conflate token frequency with per-occurrence attribution, but that is a methodological confound rather than a circular reduction of the central claim. No self-citation chain is used to justify the central premise, and the retrieval-effectiveness gains in Table 1 are independently reproduced from existing GPL results. The paper's central contribution is an interpretability protocol applied to existing models, so the derivation chain is self-contained and not circular.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper introduces no free parameters and no new physical or conceptual entities. Its central method depends on three axiomatic or heuristic choices: the validity of Integrated Gradients as an attribution method, the use of [PAD] as an empty baseline, and the aggregation over top-25 documents. These are methodological assumptions rather than fitted values, and the first is inherited from prior work while the latter two are introduced for this paper.

assumptions (3)

standard math Integrated Gradients axioms (sensitivity, implementation invariance, completeness) hold for the dense retriever and are a valid source of token attributions.
Invoked in Section 3 Methodology; the method inherits the IG framework of Sundararajan et al. [14].
domain assumption Replacing query or document tokens with [PAD] tokens creates an 'empty signal' baseline suitable for attribution in a bi-encoder.
State in Section 3 Baseline; this choice is not independently validated against other baselines, and the correctness of the attributions depends on it.
ad hoc to paper Summing individual token attributions over the top 25 retrieved documents reflects ranking-level importance.
Stated in Section 3 Ranking Analysis; no justification or validation is provided for this aggregation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Interpretability Analysis of Domain Adapted Dense Retrievers." pith.science (2026). https://pith.science/paper/YAMWBQ2I

@misc{pith2026250114459,
  author       = {Pith},
  title        = {Pith review of: Interpretability Analysis of Domain Adapted Dense Retrievers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YAMWBQ2I}},
  note         = {Machine review of arXiv:2501.14459}
}

read the original abstract

Dense retrievers have demonstrated significant potential for neural information retrieval; however, they exhibit a lack of robustness to domain shifts, thereby limiting their efficacy in zero-shot settings across diverse domains. Previous research has investigated unsupervised domain adaptation techniques to adapt dense retrievers to target domains. However, these studies have not focused on explainability analysis to understand how such adaptations alter the model's behavior. In this paper, we propose utilizing the integrated gradients framework to develop an interpretability method that provides both instance-based and ranking-based explanations for dense retrievers. To generate these explanations, we introduce a novel baseline that reveals both query and document attributions. This method is used to analyze the effects of domain adaptation on input attributions for query and document tokens across two datasets: the financial question answering dataset (FIQA) and the biomedical information retrieval dataset (TREC-COVID). Our visualizations reveal that domain-adapted models focus more on in-domain terminology compared to non-adapted models, exemplified by terms such as "hedge," "gold," "corona," and "disease." This research addresses how unsupervised domain adaptation techniques influence the behavior of dense retrievers when adapted to new domains. Additionally, we demonstrate that integrated gradients are a viable choice for explaining and analyzing the internal mechanisms of these opaque neural models.

Figures

Figures reproduced from arXiv: 2501.14459 by the authors.

**Figure 3.** Attribution analysis for random query and relevant [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

**Figure 5.** Sum of title attribution scores for TREC-COVID [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 5 canonical work pages

[1]

Avishek Anand, Procheta Sen, Sourav Saha, Manisha Verma, and Mandar Mitra

work page
[2]

Payal Bajaj, Daniel Campos, Nick Craswell, Li Deng, Jianfeng Gao, Xiaodong Liu, Rangan Majumder, Andrew McNamara, Bhaskar Mitra, Tri Nguyen, Mir Rosenberg, Xia Song, Alina Stoica, Saurabh Tiwary, and Tong Wang. 2018. MS MARCO: A Human Generated MAchine Reading COmprehension Dataset. arXiv:1611.09268 [cs.CL]

arXiv 2018
[3]

Xilun Chen, Kushal Lakhotia, Barlas Oguz, Anchit Gupta, Patrick Lewis, Stan Peshterliev, Yashar Mehdad, Sonal Gupta, and Wen-tau Yih. 2022. Salient Phrase Aware Dense Retrieval: Can a Dense Retriever Imitate a Sparse One?. In Findings of the Association for Computational Linguistics: EMNLP 2022, Yoav Goldberg, Zornitsa Kozareva, and Yue Zhang (Eds.). Asso...

doi:10.18653/v1/2022.findings-emnlp.19 2022
[4]

Suchin Gururangan, Ana Marasović, Swabha Swayamdipta, Kyle Lo, Iz Beltagy, Doug Downey, and Noah A. Smith. 2020. Don’t Stop Pretraining: Adapt Language Models to Domains and Tasks. InProceedings of the 58th Annual Meeting of the Association for Computational Linguistics , Dan Jurafsky, Joyce Chai, Natalie Schluter, and Joel Tetreault (Eds.). Association f...

doi:10.18653/v1/2020.acl-main.740 2020
[5]

Sebastian Hofstätter, Sheng-Chieh Lin, Jheng-Hong Yang, Jimmy Lin, and Allan Hanbury. 2021. Efficiently Teaching an Effective Dense Retriever with Balanced Topic Aware Sampling. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (<conf-loc>, <city>Virtual Event</city>, <country>Canada</countr...

arXiv 2021
[6]

Kenton Lee, Ming-Wei Chang, and Kristina Toutanova. 2019. Latent Retrieval for Weakly Supervised Open Domain Question Answering. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , Anna Korhonen, David Traum, and Lluís Màrquez (Eds.). Association for Computational Linguistics, Florence, Italy, 6086–6096. https://do...

doi:10.18653/v1/p19-1612 2019
[7]

Fangyu Liu, Ivan Vulić, Anna Korhonen, and Nigel Collier. 2021. Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders. InProceedings of the 2021 Conference on Empirical Methods in Natural Language Processing , Marie-Francine Moens, Xuanjing Huang, Lucia Specia, and Scott Wen-tau Yih (Eds.). ...

doi:10.18653/v1/2021.emnlp-main 2021
[8]

Scott M Lundberg and Su-In Lee. 2017. A Unified Approach to Interpreting Model Predictions. In Advances in Neural Information Processing Systems , I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (Eds.), Vol. 30. Curran Associates, Inc. https://proceedings.neurips.cc/paper_ files/paper/2017/file/8a20a8621978632d7...

work page 2017

Show all 22 references

[9]

Ji Ma, Ivan Korotkov, Yinfei Yang, Keith Hall, and Ryan McDonald. 2021. Zero-shot Neural Passage Retrieval via Domain-targeted Synthetic Question Generation. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Vo...

2021 doi
[10]

Hall, and Ryan T

Ji Ma, Ivan Korotkov, Yinfei Yang, Keith B. Hall, and Ryan T. McDonald. 2020. Zero- shot Neural Retrieval via Domain-targeted Synthetic Query Generation.ArXiv abs/2004.14503 (2020). https://api.semanticscholar.org/CorpusID:216867871

2020 arXiv
[11]

Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. InProceedings of the 2019 Conference on Empirical Methods in Natural Language Processing . Association for Computational Linguistics. https://arxiv.org/abs/1908.10084

2019 arXiv
[12]

Why Should I Trust You?

Marco Ribeiro, Sameer Singh, and Carlos Guestrin. 2016. “Why Should I Trust You?”: Explaining the Predictions of Any Classifier. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations, John DeNero, M...

2016 doi
[13]

Avanti Shrikumar, Peyton Greenside, and Anshul Kundaje. 2017. Learning Important Features Through Propagating Activation Differences. InProceedings of the 34th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 70) , Doina Precup and Y...

2017
[14]

Mukund Sundararajan, Ankur Taly, and Qiqi Yan. 2017. Axiomatic Attribution for Deep Networks. arXiv:1703.01365 [cs.LG]

2017 arXiv
[15]

Nandan Thakur, Nils Reimers, Johannes Daxenberger, and Iryna Gurevych. 2021. Augmented SBERT: Data Augmentation Method for Improving Bi-Encoders for Pairwise Sentence Scoring Tasks. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Comp...

2021
[16]

Nandan Thakur, Nils Reimers, Andreas Rücklé, Abhishek Srivastava, and Iryna Gurevych. 2021. BEIR: A Heterogenous Benchmark for Zero-shot Evaluation of Information Retrieval Models. (4 2021). https://arxiv.org/abs/2104.08663v4

2021 arXiv
[17]

Kexin Wang, Nandan Thakur, Nils Reimers, and Iryna Gurevych. 2021. GPL: Gen- erative Pseudo Labeling for Unsupervised Domain Adaptation of Dense Retrieval. arXiv preprint arXiv:2112.07577 (4 2021). https://arxiv.org/abs/2112.07577

2021 arXiv
[18]

Lampert, Bernt Schiele, and Zeynep Akata

Yongqin Xian, Christoph H. Lampert, Bernt Schiele, and Zeynep Akata. 2019. Zero-Shot Learning—A Comprehensive Evaluation of the Good, the Bad and the Ugly. IEEE Transactions on Pattern Analysis and Machine Intelligence 41, 9 (2019), 2251–2265. https://doi.org/10.1109/TPAMI.201...

2019
[19]

Ji Xin, Chenyan Xiong, Ashwin Srinivasan, Ankita Sharma, Damien Jose, and Paul Bennett. 2022. Zero-Shot Dense Retrieval with Momentum Adversarial Domain Invariant Representations. In Findings of the Association for Computational Linguistics: ACL 2022, Smaranda Muresan, Preslav...

2022 doi
[20]

Jingtao Zhan, Jiaxin Mao, Yiqun Liu, Min Zhang, and Shaoping Ma. 2020. An Analysis of BERT in Document Ranking. InProceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (Virtual Event, China) (SIGIR ’20). Association for...

2020
[21]

Wayne Xin Zhao, Jing Liu, Ruiyang Ren, and Ji-Rong Wen. 2023. Dense Text Retrieval based on Pretrained Language Models: A Survey.ACM Trans. Inf. Syst. (dec 2023). https://doi.org/10.1145/3637870 Just Accepted

2023 doi
[2023]

Explainable Information Retrieval. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (<conf-loc>, <city>Taipei</city>, <country>Taiwan</country>, </conf-loc>) (SIGIR ’23). Association for Computing Machinery, New ...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Avishek Anand, Procheta Sen, Sourav Saha, Manisha Verma, and Mandar Mitra

work page

[2] [2]

Payal Bajaj, Daniel Campos, Nick Craswell, Li Deng, Jianfeng Gao, Xiaodong Liu, Rangan Majumder, Andrew McNamara, Bhaskar Mitra, Tri Nguyen, Mir Rosenberg, Xia Song, Alina Stoica, Saurabh Tiwary, and Tong Wang. 2018. MS MARCO: A Human Generated MAchine Reading COmprehension Dataset. arXiv:1611.09268 [cs.CL]

arXiv 2018

[3] [3]

Xilun Chen, Kushal Lakhotia, Barlas Oguz, Anchit Gupta, Patrick Lewis, Stan Peshterliev, Yashar Mehdad, Sonal Gupta, and Wen-tau Yih. 2022. Salient Phrase Aware Dense Retrieval: Can a Dense Retriever Imitate a Sparse One?. In Findings of the Association for Computational Linguistics: EMNLP 2022, Yoav Goldberg, Zornitsa Kozareva, and Yue Zhang (Eds.). Asso...

doi:10.18653/v1/2022.findings-emnlp.19 2022

[4] [4]

Suchin Gururangan, Ana Marasović, Swabha Swayamdipta, Kyle Lo, Iz Beltagy, Doug Downey, and Noah A. Smith. 2020. Don’t Stop Pretraining: Adapt Language Models to Domains and Tasks. InProceedings of the 58th Annual Meeting of the Association for Computational Linguistics , Dan Jurafsky, Joyce Chai, Natalie Schluter, and Joel Tetreault (Eds.). Association f...

doi:10.18653/v1/2020.acl-main.740 2020

[5] [5]

Sebastian Hofstätter, Sheng-Chieh Lin, Jheng-Hong Yang, Jimmy Lin, and Allan Hanbury. 2021. Efficiently Teaching an Effective Dense Retriever with Balanced Topic Aware Sampling. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (<conf-loc>, <city>Virtual Event</city>, <country>Canada</countr...

arXiv 2021

[6] [6]

Kenton Lee, Ming-Wei Chang, and Kristina Toutanova. 2019. Latent Retrieval for Weakly Supervised Open Domain Question Answering. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , Anna Korhonen, David Traum, and Lluís Màrquez (Eds.). Association for Computational Linguistics, Florence, Italy, 6086–6096. https://do...

doi:10.18653/v1/p19-1612 2019

[7] [7]

Fangyu Liu, Ivan Vulić, Anna Korhonen, and Nigel Collier. 2021. Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders. InProceedings of the 2021 Conference on Empirical Methods in Natural Language Processing , Marie-Francine Moens, Xuanjing Huang, Lucia Specia, and Scott Wen-tau Yih (Eds.). ...

doi:10.18653/v1/2021.emnlp-main 2021

[8] [8]

Scott M Lundberg and Su-In Lee. 2017. A Unified Approach to Interpreting Model Predictions. In Advances in Neural Information Processing Systems , I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (Eds.), Vol. 30. Curran Associates, Inc. https://proceedings.neurips.cc/paper_ files/paper/2017/file/8a20a8621978632d7...

work page 2017