Inducing Diversity in Differentiable Search Indexing

Abhijeet Phatak; Chittaranjan Tripathy; Jayant Sachdev; Sean D Rosario; Swati Kirti

REVIEW 5 major objections 7 minor 17 references

Inducing Diversity in Differentiable Search Indexing

T0 review · 5 major / 7 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read Diversity can be trained into a differentiable search index at no significant relevance cost.

desk verdict A plausible, clearly-written extension to DSI that doesn't yet earn its central claim: the diversity loss is computed on doc-ID prototypes, and the paper never shows those correlate with text-level diversity. read the letter →

arxiv 2502.02788 v1 pith:ITDHYBZM submitted 2025-02-05 cs.IR cs.AIcs.LG

Abhijeet Phatak , Jayant Sachdev , Sean D Rosario , Swati Kirti , Chittaranjan Tripathy This is my paper · ORCID

classification cs.IRcs.AIcs.LG

keywords DifferentiableSearchIndexInformationRetrievalRelevanceandDiversityMaximalMarginalNeuralNetworksTransformersNaturalLanguageProcessingDiversity-awaretrainingloss

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 claims that a differentiable search index (DSI) — a transformer that maps a query directly to document identifiers instead of scanning a separate index — can be trained to return diverse results by adding a redundancy penalty to its loss. The proposed loss combines the usual cross-entropy relevance term with a weighted sum of cosine similarities among the top-$K$ document embeddings, inspired by Maximal Marginal Relevance. On NQ320K and MSMARCO, models trained with this extra term keep the ranking-accuracy metrics (Hits@K and MRR@10) close to the naive DSI baseline while reducing ROUGE-L homogeneity and increasing n-gram diversity in the retrieved set. If the claim holds, a single end-to-end training step can replace the common post-processing diversification step, with no extra inference latency.

What carries the argument

The load-bearing object is the modified loss together with the representation it acts on. In this DSI variant, a transformer classifier scores all document identifiers, and the paper takes the top-$K$ logits, extracts the document embedding vectors behind them from the final linear layer, and adds their pairwise cosine similarity $\mathrm{Sim}_2$ to the training objective. This term penalizes the model whenever the top-$K$ documents it predicts for a query are mutually similar, so training spreads those high-probability vectors apart. The MMR-inspired combination is $\mathcal{L}_{\rm total} = \alpha \cdot \mathrm{CE} + (1-\alpha) \sum_{d_i,d_j \in K, j>i} \mathrm{Sim}_2(d_i,d_j)$, with the cross-entropy term $\mathrm{CE}$ carrying relevance and $\alpha$ balancing the two.

What would settle it

On a held-out set of queries, compute the rank correlation between the pairwise cosine similarities of the top-K document vectors and the pairwise ROUGE-L similarities of the corresponding documents; if the correlation is near zero (or negative on queries with highly repetitive outputs), the diversity loss is not optimizing the redundancy it reports to reduce.

Watch

Extended reading notes

Core claim

The central claim is that diversity can be induced inside the DSI training objective rather than applied at retrieval time. With a weighting $\alpha < 1$, the total loss $\mathcal{L}_{\rm total} = \alpha \cdot \mathrm{CE} + (1-\alpha) \sum_{d_i,d_j \in K, j>i} \mathrm{Sim}_2(d_i,d_j)$ pushes the top-$K$ document vectors apart during training. The resulting model returns sets of documents that score lower on ROUGE-L homogeneity and higher on n-gram diversity while matching the baseline on Hits@1/5/10 and MRR@10. The paper reads this as evidence that DSI can learn to diversify while staying relevant, removing the need for a post-processing diversification stage.

Load-bearing premise

The method assumes that the cosine similarity between the learned document vectors in the final classifier layer is a faithful proxy for how similar or redundant the actual retrieved documents are.

Editorial extensions

If this is right

With $\alpha < 1$, the trained DSI returns top-$K$ sets with lower ROUGE-L homogeneity and higher n-gram diversity than the naive DSI baseline on both NQ320K and MSMARCO.
Hits@1/5/10 and MRR@10 stay close to (and in several rows slightly above) the naive baseline, so the diversity gain does not come at a substantial relevance cost.
Because diversification is encoded in the model weights, no MMR-style post-processing or beam-search diversification is required at query time, and inference latency is unchanged.
The parameter $\alpha$ provides a single tuning knob for the relevance--diversity trade-off, and the approach extends to incremental DSI settings where the index is updated with new documents.

Reading between the lines

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

A testable extension the paper does not explore: check whether the cosine similarity of the top-$K$ classifier embeddings actually correlates with pairwise ROUGE-L similarity of the corresponding documents; if the correlation is weak, the loss may be optimizing a proxy that does not generalize to genuinely diverse outputs.
The same redundancy penalty could apply to other neural retrieval models that score documents with embedding vectors, and to generative DSI variants, by applying $\mathrm{Sim}_2$ to the representation the decoder uses.
The loss averages similarity over all top-$K$ pairs, whereas MMR penalizes similarity only to already-selected documents; an order-aware or max-based variant might give more stable diversity for larger $K$.
Sub-topic-coverage judgments, rather than lexical proxy metrics such as ROUGE-L and NGD, would reveal whether the induced diversity actually helps users find distinct aspects of a query.

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

Quick take: this is a reasonable, small idea - add an MMR-style diversity penalty to the DSI training loss - and the writing is honest. But the headline claim ('diversity without significant impact to relevance') rests on a proxy that is never validated, and the reported diversity metrics are inconsistent. I'd send it to review, but conditional on fixing that gap.

What's new: nobody has put a diversity term directly into DSI training. The loss in Section 4.2 is simple: cross-entropy plus a cosine-similarity penalty on the top-K doc-ID logits, interpolated by alpha. That's a legitimate extension, and the authors are careful to note that this removes the need for post-processing MMR at inference. The qualitative example for 'California' is useful, and it's good that they report inference time.

What's weak: the paper never tests whether cosine similarity in the final-layer document-ID embedding space tracks textual redundancy. Those vectors are class prototypes trained to map queries to labels; their geometry may be unrelated to ROUGE-L, n-gram diversity, or compression ratio. The stress-test note is right: this is the load-bearing assumption. Also, the results are mixed. On MSMARCO, alpha=0.25 gives worse ROUGE-L than baseline, and alpha=0.75 gives worse compression ratio. On NQ320K the diversity gains are more consistent, but there are no error bars, no multiple seeds, no significance tests, and the training-time K is never reported. For a conference paper that's incomplete; for a workshop paper it's borderline.

The non-circularity is a plus: the evaluation uses external text metrics, so the reported improvements aren't generated by construction from the loss. But external metrics can only help if the loss is actually optimizing a good proxy, and that link is missing.

Who should read it: people working on DSI or differentiable retrieval, especially IncDSI. They'd see this as an obvious next step and want to know if it holds. The paper is short, easy to follow, and the authors are clear about their setup.

Recommendation: send to peer review with a request for proxy validation (correlate Sim2 with ROUGE-L/CR on held-out pairs), multiple seeds, and a stated K. The idea deserves a serious referee; the current evidence doesn't support the central claim as written.

Referee Report

5 major / 7 minor

Summary. The paper proposes a training-time method to induce diversity in Differentiable Search Indexing (DSI). The authors modify the standard cross-entropy training loss by adding a regularizer that penalizes cosine similarity among the document-ID embedding vectors corresponding to the top-K predicted logits, inspired by Maximal Marginal Relevance (MMR). The total loss is a convex combination of cross-entropy and this pairwise similarity penalty, weighted by a parameter alpha. They evaluate on NQ320K and MSMARCO for alpha values 0.25, 0.5, 0.75, and 1.0, reporting relevance metrics (Hits@K, MRR@10) and diversity metrics (ROUGE-L homogeneity, n-gram diversity, compression ratio). The authors claim that alpha < 1 yields more diverse retrieval sets with no significant loss in relevance, and provide a qualitative example for the query 'California'.

Significance. If validated, this is a simple and efficient way to inject diversity into DSI training without requiring post-processing or extra latency at inference time. The proposed loss modification is easy to implement and the paper evaluates it on two standard datasets with multiple diversity metrics. The qualitative example is illustrative. However, the empirical support is currently incomplete: there is no validation that the cosine similarity in the document-ID embedding space actually tracks the text-based diversity metrics used for evaluation, no comparison against an MMR post-processing baseline, and no statistical reliability checks. The central claim is therefore plausible but not yet convincingly established.

major comments (5)

[Section 4.2, loss formula] The diversity regularizer penalizes cosine similarity among document-ID embedding vectors in the final classification layer, while the evaluation metrics (ROUGE-L, NGD, CR) are computed on the text of retrieved documents. The paper does not report any correlation between pairwise cosine similarity in the doc-ID embedding space and the text-based diversity metrics. The authors should validate this proxy, for example by computing the correlation on the retrieved sets or by ablating with a diversity term computed directly on text embeddings (e.g., DistilBERT sentence embeddings). Without such validation, it is unclear whether the observed diversity improvements are caused by the intended mechanism or by unintended changes to the classifier geometry.
[Table 2, Section 5] The claim that diversity is achieved 'without any significant impact to relevance' is not supported for MSMARCO. For alpha=0.75, Hits@1 drops from 0.4837 to 0.4664, Hits@5 from 0.7419 to 0.7240, and MRR@10 from 0.5937 to 0.5757. In addition, diversity metrics move inconsistently across alpha: on MSMARCO alpha=0.25 gives ROUGE-L 0.188 versus 0.165 for alpha=1 (i.e., less diverse by this metric), while alpha=0.75 gives CR 1.242 versus 1.204 (also less diverse). The paper reports single runs without error bars or significance tests, so it is impossible to assess whether these differences are meaningful. Please report means and standard deviations over multiple seeds and perform significance tests for both relevance and diversity metrics.
[Section 4.3] The value of K used in the diversity loss is never stated. The loss in Section 4.2 depends on K both through the number of pairwise terms and the composition of the selected top-K set, and the results are likely sensitive to this hyperparameter. Without reporting K, the experiments are not reproducible. Please specify K and include a sensitivity analysis over K (e.g., K=5, 10, 20).
[Sections 1 and 6] The paper motivates the approach as eliminating the need for MMR post-processing, but no experiment compares the proposed method against an MMR baseline applied to naive DSI outputs. To support the claim that end-to-end diversity training is preferable to post-processing, the authors should include an MMR post-processing baseline (e.g., with several lambda values) and compare the relevance-diversity tradeoff with the proposed method.
[Section 4.2] The diversity loss is computed over the top-K predicted documents, but the top-K selection is a non-differentiable operation. The paper does not describe how this is handled during backpropagation (e.g., straight-through estimation or treating the selection as fixed). The authors should clarify this, as it affects the validity of the gradient updates and the interpretation of the training procedure.

minor comments (7)

[Section 4.2, loss formula] The notation 'di,dj ∈ K, j>i' is ambiguous because K is also used as an integer in Hits@K. Please use a different symbol for the set of top-K documents (e.g., S_K) and clarify whether the sum is normalized by the number of pairs.
[Table 1] The description of ROUGE-L says the similarity is 'computed using uncased DistilBERT embeddings,' which is not standard ROUGE-L (usually based on longest common subsequence). Please clarify the exact computation.
[Section 4.3] The word 'sigle' should be 'single'.
[Section 5] The phrase 'the diversity-induced models slightly outperform the naive DSI model' is not accurate for MSMARCO alpha=0.75, where relevance metrics are lower; please rephrase to reflect the actual results.
[Table 3] The qualitative analysis is based on a single query; consider adding more examples or a small structured analysis to strengthen the claim.
[Abstract and Conclusion] The abstract and conclusion state the method 'can also be easily be extended to the incremental DSI settings' (typo: 'be easily be'); since no experiments are presented, please soften this claim or add a discussion of the expected behavior.
[Section 4.1] The paper does not mention the source of the IncDSI data splits or provide URLs; please add details for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the diversity loss is evaluated against independent text-based metrics, and the trained model's behavior is empirically measured rather than forced by construction.

full rationale

The paper's claimed contribution is a modified training loss, Losstotal = alpha * cross-entropy + (1-alpha) * sum of Sim2 over top-K document ID embedding vectors (Section 4.2), and the claim is that this induces diversity without hurting relevance. The evaluation uses Hits@K, MRR@10, ROUGE-L, NGD, and CR computed on the actual retrieved document texts, which are external to and not identical with the cosine-similarity term optimized during training. There is no equation in the paper that reduces the reported diversity metrics to the training loss by construction, nor any fitted parameter that is renamed as a prediction. The alpha grid is a hyperparameter sweep, not a fitted quantity used to define the target. The cited IncDSI work is used only to supply the baseline DSI implementation and data splits; it is not a self-citation and does not carry the diversity claim. The central validity concern—whether cosine similarity in the classifier's final linear layer faithfully tracks textual redundancy—is an empirical proxy-validation question, not a circularity: if the proxy fails, the method would simply be ineffective, not tautological. Therefore no load-bearing circular step is identifiable, and the paper is best assessed as self-contained empirical work whose limitations are about proxy validity rather than circular reasoning.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The method introduces no new entities. The free parameters are alpha and K. The main unstated premise is that document ID vector similarity is a proxy for textual diversity.

free parameters (2)

alpha = grid {1, 0.75, 0.5, 0.25}
Balances the classification loss and the diversity term in Section 4.2. The paper reports results for each value but does not justify a single best value.
K (top-K logits) = not stated
The diversity term in Section 4.2 is computed over the top K predicted logits, but K is never specified, making the method under-specified.

assumptions (3)

domain assumption Cosine similarity among classifier document ID embedding vectors reflects semantic or textual diversity of the corresponding documents.
Section 4.2 uses the linear layer's top-K vectors as Sim2 without validating the correlation with text similarity.
domain assumption The top-K selection is differentiable or can be trained through with a surrogate; the paper does not discuss how gradients are computed through the non-differentiable argmax selection.
Required for the loss in Section 4.2 to train the model, but unstated.
domain assumption The IncDSI data splits are appropriate for evaluating the proposed method.
Section 4.1 says they use the IncDSI splits; any leak or mismatch would affect results.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Inducing Diversity in Differentiable Search Indexing." pith.science (2026). https://pith.science/paper/ITDHYBZM

@misc{pith2026250202788,
  author       = {Pith},
  title        = {Pith review of: Inducing Diversity in Differentiable Search Indexing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ITDHYBZM}},
  note         = {Machine review of arXiv:2502.02788}
}

read the original abstract

Differentiable Search Indexing (DSI) is a recent paradigm for information retrieval which uses a transformer-based neural network architecture as the document index to simplify the retrieval process. A differentiable index has many advantages enabling modifications, updates or extensions to the index. In this work, we explore balancing relevance and novel information content (diversity) for training DSI systems inspired by Maximal Marginal Relevance (MMR), and show the benefits of our approach over the naive DSI training. We present quantitative and qualitative evaluations of relevance and diversity measures obtained using our method on NQ320K and MSMARCO datasets in comparison to naive DSI. With our approach, it is possible to achieve diversity without any significant impact to relevance. Since we induce diversity while training DSI, the trained model has learned to diversify while being relevant. This obviates the need for a post-processing step to induce diversity in the recall set as typically performed using MMR. Our approach will be useful for Information Retrieval problems where both relevance and diversity are important such as in sub-topic retrieval. Our work can also be easily be extended to the incremental DSI settings which would enable fast updates to the index while retrieving a diverse recall set.

Figures

Figures reproduced from arXiv: 2502.02788 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 12 canonical work pages

[1]

In: Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval

Carbonell, J., Goldstein, J.: The use of mmr, diversity-based reranking for re- ordering documents and producing summaries. In: Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval. pp. 335–336 (1998)

work page 1998
[2]

In: North American Chap- ter of the Association for Computational Linguistics (2019), https://api

Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: Bert: Pre-training of deep bidi- rectional transformers for language understanding. In: North American Chap- ter of the Association for Computational Linguistics (2019), https://api. semanticscholar.org/CorpusID:52967399

work page 2019
[3]

In: The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval

Feng, Y., Xu, J., Lan, Y., Guo, J., Zeng, W., Cheng, X.: From greedy selection to exploratory decision-making: Diverse ranking with policy-value networks. In: The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval. p. 125–134. SIGIR ’18, Association for Computing Machin- ery, New York, NY, USA (2018). https://doi.or...

arXiv 2018
[4]

In: Proceed- ingsofthe33rdinternationalACMSIGIRconferenceonResearchanddevelopment in information retrieval

Guo, S., Sanner, S.: Probabilistic latent maximal marginal relevance. In: Proceed- ingsofthe33rdinternationalACMSIGIRconferenceonResearchanddevelopment in information retrieval. pp. 833–834 (2010)

work page 2010
[5]

In: International Conference on Machine Learning

Kishore, V., Wan, C., Lovelace, J., Artzi, Y., Weinberger, K.Q.: Incdsi: incre- mentally updatable document retrieval. In: International Conference on Machine Learning. pp. 17122–17134. PMLR (2023)

work page 2023
[6]

Transactions of the Association for Compu- tational Linguistics 7, 453–466 (2019)

Kwiatkowski, T., Palomaki, J., Redfield, O., Collins, M., Parikh, A., Alberti, C., Epstein, D., Polosukhin, I., Devlin, J., Lee, K., et al.: Natural questions: a bench- mark for question answering research. Transactions of the Association for Compu- tational Linguistics 7, 453–466 (2019)

2019
[7]

In: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval

Liu, J., Dou, Z., Wang, X., Lu, S., Wen, J.R.: Dvgan: A minimax game for search result diversification combining explicit and implicit features. In: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. pp. 479–488 (2020)

work page 2020
[8]

arXiv preprint arXiv:2212.09744 (2022)

Mehta, S.V., Gupta, J., Tay, Y., Dehghani, M., Tran, V.Q., Rao, J., Najork, M., Strubell, E., Metzler, D.: Dsi++: Updating transformer memory with new docu- ments. arXiv preprint arXiv:2212.09744 (2022)

arXiv 2022

Show all 17 references

[9]

arXiv preprint arXiv:1611.09268 (2016)

Nguyen, T.: Ms marco: A human generated machine reading comprehension dataset. arXiv preprint arXiv:1611.09268 (2016)

2016 arXiv
[10]

Advances in Neural Information Processing Systems35, 21831–21843 (2022)

Tay, Y., Tran, V., Dehghani, M., Ni, J., Bahri, D., Mehta, H., Qin, Z., Hui, K., Zhao, Z., Gupta, J., et al.: Transformer memory as a differentiable search index. Advances in Neural Information Processing Systems35, 21831–21843 (2022)

2022
[11]

Enterprise Information Systems17(5), 1992018 (2023)

Wu, C.H., Wang, Y., Ma, J.: Maximal marginal relevance-based recommendation for product customisation. Enterprise Information Systems17(5), 1992018 (2023)

2023
[12]

IEEE Trans

Wu,H.,Zhang,Y.,Ma,C.,Lyu,F.,He,B.,Mitra,B.,Liu,X.:Resultdiversification in search and recommendation: A survey. IEEE Trans. on Knowl. and Data Eng. 36(10), 5354–5373 (Apr 2024). https://doi.org/10.1109/TKDE.2024.3382262, https://doi.org/10.1109/TKDE.2024.3382262

2024
[13]

In: Proceedings of the 38th international ACM SIGIR conference on research and development in information retrieval

Xia, L., Xu, J., Lan, Y., Guo, J., Cheng, X.: Learning maximal marginal relevance model via directly optimizing diversity evaluation measures. In: Proceedings of the 38th international ACM SIGIR conference on research and development in information retrieval. pp. 113–122 (2015)

2015
[14]

In: Proceedings of the 40th International 8 A

Xia, L., Xu, J., Lan, Y., Guo, J., Zeng, W., Cheng, X.: Adapting markov decision process for search result diversification. In: Proceedings of the 40th International 8 A. Phatak et al. ACM SIGIR Conference on Research and Development in Information Retrieval. p. 535–544. SIGIR...

2017
[15]

In: Proceedings of the Web Conference 2021

Yan, L., Qin, Z., Pasumarthi, R.K., Wang, X., Bendersky, M.: Diversification- aware learning to rank using distributed representation. In: Proceedings of the Web Conference 2021. p. 127–136. WWW ’21, Association for Computing Machin- ery, New York, NY, USA (2021). https://doi....

2021
[16]

arXiv preprint arXiv:2211.13892 (2022)

Ye, X., Iyer, S., Celikyilmaz, A., Stoyanov, V., Durrett, G., Pasunuru, R.: Complementary explanations for effective in-context learning. arXiv preprint arXiv:2211.13892 (2022)

2022 arXiv
[17]

In: Proceedings of the 37th International ACM SIGIR Conference on Research & Development in Information Retrieval

Zhu, Y., Lan, Y., Guo, J., Cheng, X., Niu, S.: Learning for search result diver- sification. In: Proceedings of the 37th International ACM SIGIR Conference on Research & Development in Information Retrieval. p. 293–302. SIGIR ’14, Associ- ation for Computing Machinery, New Yor...

2014

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.

[1] [1]

In: Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval

Carbonell, J., Goldstein, J.: The use of mmr, diversity-based reranking for re- ordering documents and producing summaries. In: Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval. pp. 335–336 (1998)

work page 1998

[2] [2]

In: North American Chap- ter of the Association for Computational Linguistics (2019), https://api

Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: Bert: Pre-training of deep bidi- rectional transformers for language understanding. In: North American Chap- ter of the Association for Computational Linguistics (2019), https://api. semanticscholar.org/CorpusID:52967399

work page 2019

[3] [3]

In: The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval

Feng, Y., Xu, J., Lan, Y., Guo, J., Zeng, W., Cheng, X.: From greedy selection to exploratory decision-making: Diverse ranking with policy-value networks. In: The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval. p. 125–134. SIGIR ’18, Association for Computing Machin- ery, New York, NY, USA (2018). https://doi.or...

arXiv 2018

[4] [4]

In: Proceed- ingsofthe33rdinternationalACMSIGIRconferenceonResearchanddevelopment in information retrieval

Guo, S., Sanner, S.: Probabilistic latent maximal marginal relevance. In: Proceed- ingsofthe33rdinternationalACMSIGIRconferenceonResearchanddevelopment in information retrieval. pp. 833–834 (2010)

work page 2010

[5] [5]

In: International Conference on Machine Learning

Kishore, V., Wan, C., Lovelace, J., Artzi, Y., Weinberger, K.Q.: Incdsi: incre- mentally updatable document retrieval. In: International Conference on Machine Learning. pp. 17122–17134. PMLR (2023)

work page 2023

[6] [6]

Transactions of the Association for Compu- tational Linguistics 7, 453–466 (2019)

Kwiatkowski, T., Palomaki, J., Redfield, O., Collins, M., Parikh, A., Alberti, C., Epstein, D., Polosukhin, I., Devlin, J., Lee, K., et al.: Natural questions: a bench- mark for question answering research. Transactions of the Association for Compu- tational Linguistics 7, 453–466 (2019)

2019

[7] [7]

In: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval

Liu, J., Dou, Z., Wang, X., Lu, S., Wen, J.R.: Dvgan: A minimax game for search result diversification combining explicit and implicit features. In: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. pp. 479–488 (2020)

work page 2020

[8] [8]

arXiv preprint arXiv:2212.09744 (2022)

Mehta, S.V., Gupta, J., Tay, Y., Dehghani, M., Tran, V.Q., Rao, J., Najork, M., Strubell, E., Metzler, D.: Dsi++: Updating transformer memory with new docu- ments. arXiv preprint arXiv:2212.09744 (2022)

arXiv 2022