pith. sign in

arxiv: 2606.17468 · v1 · pith:7HVQSBSBnew · submitted 2026-06-16 · 💻 cs.IR

RSRank: Learning Relevance from Representational Shifts

Archit Gupta , Sai Sundaresan , Debabrata Mahapatra This is my paper

Pith reviewed 2026-06-26 23:03 UTC · model grok-4.3

classification 💻 cs.IR
keywords rerankingrepresentational shiftsrelevance scoringRAG systemsinformation retrievallanguage model internalscalibrated scoring
0
0 comments X

The pith

The alignment between representational shifts induced by a candidate document and those from an oracle document set indicates relevance for reranking.

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

The paper establishes that representational shifts in a language model's internal query state when conditioned on a document provide a signal for relevance. This signal is measured by how closely the shift from a candidate document matches the shift from an oracle set of relevant documents. A lightweight training framework then projects these alignments into calibrated relevance scores. The method avoids next-token prediction logits and reduces reliance on heuristic thresholds for filtering. Experiments show gains over existing rerankers on diverse retrieval datasets.

Core claim

We identify a principled signal for relevance: the representational shift (RS) induced in a query's internal state when conditioned on a document. We observe that the alignment between (a) RS induced by a candidate document and (b) RS induced by an oracle document-set provides a robust indicator of relevance. Building on this insight, we introduce a lightweight training framework that learns projections mapping RS to calibrated relevance scores. Our training objectives naturally filter irrelevant content at a zero threshold, reducing dependence on heuristic tuning. Across diverse retrieval datasets, our method delivers gains over SOTA rerankers.

What carries the argument

Representational shift (RS) — the change in a query's internal state when conditioned on a document — with alignment to oracle-induced RS as the relevance indicator.

If this is right

  • RS alignment supplies a relevance signal independent of next-token logits.
  • Training projects RS values to scores that separate relevant from irrelevant content at a zero threshold.
  • The approach reduces dependence on manual heuristic threshold selection in RAG rerankers.
  • Performance improves over state-of-the-art rerankers on multiple retrieval datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same RS alignment could be tested as a relevance signal in non-RAG retrieval settings where internal states remain accessible.
  • If the signal holds across model scales, the method might support reranking without requiring logit access or full model fine-tuning.
  • The zero-threshold property might simplify deployment in production systems that must handle varying query distributions.
  • Connections to other internal-state analyses in language models could be explored to see whether RS alignment generalizes beyond reranking.

Load-bearing premise

Representational shifts induced in a query's internal state when conditioned on documents form a principled and generalizable signal for relevance that can be projected to calibrated scores via lightweight training without dataset-specific overfitting.

What would settle it

On a held-out retrieval dataset, if alignment scores between candidate-document RS and oracle-set RS show no higher correlation with human relevance judgments than logit-based baselines, the central claim would be falsified.

Figures

Figures reproduced from arXiv: 2606.17468 by Archit Gupta, Debabrata Mahapatra, Sai Sundaresan.

Figure 1
Figure 1. Figure 1: Optimal threshold for Qwen3-Reranker-8B for F1 across datasets. The x-axis shows the range of scores (globally normalized); the optimal threshold is indicated by the red dot. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Per-Query F1 Gap 0 20 40 60 80 100 % of Queries Exceeding Gap 63% 40% 30% [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: F1 gap CDF for Qwen3-Reranker-8B on HotpotQA. The x-axis shows the F1 gap between the dataset-level optimal threshold and the per-query opti￾mal threshold; the y-axis shows the fraction of queries exceeding that gap. 63% of queries lose ≥0.1 F1 from using a fixed threshold. To quantify performance loss attributable specifi￾cally to poor calibration rather than reranking qual￾ity, we conduct a paired t-test… view at source ↗
Figure 4
Figure 4. Figure 4: UMAP visualization of representational shifts [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Optimal threshold for RSRank for best mean [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: decomposes each model’s F1 into two components: the dataset-optimal F1 and the additional headroom to per-query optimal F1. RSRank achieves a higher per-query optimal F1 than Qwen3-Reranker-8B on average, indicating stronger underlying ranking quality. The headroom from dataset-optimal to query-optimal is larger for RSRank (+15.9 vs. +12.6 on average). These re￾sults indicate that RSRank provides a strong … view at source ↗
read the original abstract

As enterprises deploy RAG-based systems to provide grounded responses to user queries, reranking has become a critical component for the final filtering step that separates relevant from distracting or irrelevant documents. Existing rerankers often rely on heuristic thresholds to achieve optimal filtering. Moreover, for relevance scoring, state-of-the-art methods use a language model's logit signals, which are designed for next-token prediction, not for assessing relevance. To address these limitations, we identify a principled signal for relevance: the representational shift (RS) induced in a query's internal state when conditioned on a document. We observe that the alignment between (a) RS induced by a candidate document and (b) RS induced by an oracle document-set provides a robust indicator of relevance. Building on this insight, we introduce a lightweight training framework that learns projections mapping RS to calibrated relevance scores. Our training objectives naturally filter irrelevant content at a zero threshold, reducing dependence on heuristic tuning. Across diverse retrieval datasets, our method delivers gains over SOTA rerankers.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The manuscript introduces RSRank for reranking in RAG systems. It identifies the representational shift (RS) induced in a query's internal state when conditioned on a document as a relevance signal, observing that alignment between the RS induced by a candidate document and the RS induced by an oracle document-set provides a robust indicator of relevance. A lightweight training framework learns projections mapping these RS vectors to calibrated relevance scores; the training objectives are described as naturally enabling zero-threshold filtering of irrelevant content. The method is reported to deliver gains over state-of-the-art rerankers across diverse retrieval datasets.

Significance. If the empirical observations and gains hold under full experimental scrutiny, the work supplies a relevance signal grounded in internal model states rather than next-token logits, together with a training procedure that reduces dependence on post-hoc thresholds. This could strengthen the final filtering step in enterprise RAG pipelines and offers a concrete alternative to existing logit-based rerankers.

minor comments (2)
  1. [Abstract] The abstract states that the method 'delivers gains over SOTA rerankers' but supplies no quantitative deltas, dataset names, or baseline comparisons; these must appear with error bars and ablation results in §4 or §5 to allow verification of the central claim.
  2. [§3] The description of the oracle document-set construction and the precise definition of 'alignment' between RS vectors should be expanded with an equation or pseudocode in §3 to make the signal reproducible.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary of our work, the assessment of its potential significance for RAG pipelines, and the recommendation for minor revision. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's central claim rests on an empirical observation that alignment between representational shifts (RS) from candidate documents and an oracle document-set serves as a relevance signal, followed by a lightweight training framework to project RS vectors to calibrated scores. No equations, self-definitional constructions, fitted parameters renamed as predictions, or load-bearing self-citations are present in the provided abstract or description. The derivation does not reduce to its inputs by construction; the training objectives are described as naturally producing zero-threshold filtering without evidence of statistical forcing or ansatz smuggling. This is a self-contained empirical approach against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no extractable free parameters, axioms, or invented entities; full paper would be required to populate the ledger.

pith-pipeline@v0.9.1-grok · 5706 in / 1044 out tokens · 34500 ms · 2026-06-26T23:03:40.236211+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

62 extracted references · 26 canonical work pages · 1 internal anchor

  1. [1]

    Robertson, Stephen and Zaragoza, Hugo , title =. Found. Trends Inf. Retr. , month = apr, pages =. 2009 , issue_date =. doi:10.1561/1500000019 , abstract =

    work page doi:10.1561/1500000019 2009

  2. [2]

    2026 , eprint=

    LLM2Vec-Gen: Generative Embeddings from Large Language Models , author=. 2026 , eprint=

    2026

  3. [3]

    2024 , eprint=

    LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders , author=. 2024 , eprint=

    2024

  4. [4]

    Proceedings of the ACM Web Conference 2025 , year =

    Chen, Yiqun and Liu, Qi and Zhang, Yi and Sun, Weiwei and Ma, Xinyu and Yang, Weiwei and Shi, Daiting and Mao, Jiaxin and Yin, Dawei , title =. Proceedings of the ACM Web Conference 2025 , year =. doi:10.1145/3696410.3714863 , publisher =

    work page doi:10.1145/3696410.3714863 2025

  5. [5]

    arXiv preprint arXiv:2312.02724 , year=

    RankZephyr: Effective and Robust Zero-Shot Listwise Reranking is a Breeze! , author=. arXiv preprint arXiv:2312.02724 , year=

    Pith/arXiv arXiv

  6. [6]

    Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =

    Zhuang, Shengyao and Zhuang, Honglei and Koopman, Bevan and Zuccon, Guido , title =. Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =. 2024 , isbn =. doi:10.1145/3626772.3657813 , abstract =

    work page doi:10.1145/3626772.3657813 2024

  7. [7]

    Attention-Guided Hierarchical Defense for Multimodal Attacks in Vision-Language Models , year=

    Chen, Long and Chen, Yuling and Luo, Yun and Dou, Hui and Zhong, Xinyang , booktitle=. Attention-Guided Hierarchical Defense for Multimodal Attacks in Vision-Language Models , year=

  8. [8]

    Peering into the Mind of Language Models: An Approach for Attribution in Contextual Question Answering

    Phukan, Anirudh and Somasundaram, Shwetha and Saxena, Apoorv and Goswami, Koustava and Srinivasan, Balaji Vasan. Peering into the Mind of Language Models: An Approach for Attribution in Contextual Question Answering. Findings of the Association for Computational Linguistics: ACL 2024. 2024. doi:10.18653/v1/2024.findings-acl.682

    work page doi:10.18653/v1/2024.findings-acl.682 2024

  9. [9]

    2024 , url=

    Jianyi Zhang and Da-Cheng Juan and Cyrus Rashtchian and Chun-Sung Ferng and Heinrich Jiang and Yiran Chen , booktitle=. 2024 , url=

    2024

  10. [10]

    Document Re-Ranking With Evidential Neural Networks , year=

    Yoon, Jeongnoh and Sael, Lee , journal=. Document Re-Ranking With Evidential Neural Networks , year=

  11. [11]

    CoRR , volume=

    Le Zhang and Bo Wang and Xipeng Qiu and Siva Reddy and Aishwarya Agrawal , title=. CoRR , volume=. 2025 , month=

    2025

  12. [12]

    On the Sentence Embeddings from Pre-trained Language Models

    Li, Bohan and Zhou, Hao and He, Junxian and Wang, Mingxuan and Yang, Yiming and Li, Lei. On the Sentence Embeddings from Pre-trained Language Models. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020. doi:10.18653/v1/2020.emnlp-main.733

    work page doi:10.18653/v1/2020.emnlp-main.733 2020

  13. [13]

    Proceedings of the ACM on Web Conference 2025 , pages =

    Ren, Ruiyang and Wang, Yuhao and Zhou, Kun and Zhao, Wayne Xin and Wang, Wenjie and Liu, Jing and Wen, Ji-Rong and Chua, Tat-Seng , title =. Proceedings of the ACM on Web Conference 2025 , pages =. 2025 , isbn =. doi:10.1145/3696410.3714658 , abstract =

    work page doi:10.1145/3696410.3714658 2025

  14. [14]

    How Contextual are Contextualized Word Representations? C omparing the Geometry of BERT , ELM o, and GPT -2 Embeddings

    Ethayarajh, Kawin. How Contextual are Contextualized Word Representations? C omparing the Geometry of BERT , ELM o, and GPT -2 Embeddings. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019. doi:10.18653/v1/D19-1006

    work page doi:10.18653/v1/d19-1006 2019

  15. [15]

    and Jaimes, Alejandro

    Yu, Puxuan and Cohen, Daniel and Lamba, Hemank and Tetreault, Joel R. and Jaimes, Alejandro. Explain then Rank: Scale Calibration of Neural Rankers Using Natural Language Explanations from LLM s. Findings of the Association for Computational Linguistics: ACL 2025. 2025. doi:10.18653/v1/2025.findings-acl.1167

    work page doi:10.18653/v1/2025.findings-acl.1167 2025

  16. [16]

    CoRR , volume=

    Wenhan Liu and Xinyu Ma and Weiwei Sun and Yutao Zhu and Yuchen Li and Dawei Yin and Zhicheng Dou , title=. CoRR , volume=. 2025 , month=

    2025

  17. [17]

    Relevance Scores Calibration for Ranked List Truncation via TMP Adapter

    Posokhov, Pavel and Masliukhin, Sergei and Stepan, Skrylnikov and Tirskikh, Danil and Makhnytkina, Olesia. Relevance Scores Calibration for Ranked List Truncation via TMP Adapter. Findings of the Association for Computational Linguistics: ACL 2025. 2025. doi:10.18653/v1/2025.findings-acl.402

    work page doi:10.18653/v1/2025.findings-acl.402 2025

  18. [18]

    The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

    AcuRank: Uncertainty-Aware Adaptive Computation for Listwise Reranking , author=. The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

  19. [19]

    Certified Error Control of Candidate Set Pruning for Two-Stage Relevance Ranking

    Li, Minghan and Zhang, Xinyu and Xin, Ji and Zhang, Hongyang and Lin, Jimmy. Certified Error Control of Candidate Set Pruning for Two-Stage Relevance Ranking. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022. doi:10.18653/v1/2022.emnlp-main.23

    work page doi:10.18653/v1/2022.emnlp-main.23 2022

  20. [20]

    Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2) , year=

    Nandan Thakur and Nils Reimers and Andreas R. Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2) , year=

  21. [21]

    CoRR , volume=

    Ronak Pradeep and Rodrigo Frassetto Nogueira and Jimmy Lin , title=. CoRR , volume=. 2021 , cdate=

    2021

  22. [22]

    Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =

    Khattab, Omar and Zaharia, Matei , title =. Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =. 2020 , isbn =. doi:10.1145/3397271.3401075 , abstract =

    work page doi:10.1145/3397271.3401075 2020

  23. [23]

    Document Ranking with a Pretrained Sequence-to-Sequence Model

    Nogueira, Rodrigo and Jiang, Zhiying and Pradeep, Ronak and Lin, Jimmy. Document Ranking with a Pretrained Sequence-to-Sequence Model. Findings of the Association for Computational Linguistics: EMNLP 2020. 2020. doi:10.18653/v1/2020.findings-emnlp.63

    work page doi:10.18653/v1/2020.findings-emnlp.63 2020

  24. [24]

    Sparse, Dense, and Attentional Representations for Text Retrieval

    Luan, Yi and Eisenstein, Jacob and Toutanova, Kristina and Collins, Michael. Sparse, Dense, and Attentional Representations for Text Retrieval. Transactions of the Association for Computational Linguistics. 2021. doi:10.1162/tacl_a_00369

    work page doi:10.1162/tacl_a_00369 2021

  25. [25]

    Proceedings of the 34th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =

    Wang, Lidan and Lin, Jimmy and Metzler, Donald , title =. Proceedings of the 34th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =. 2011 , isbn =. doi:10.1145/2009916.2009934 , abstract =

    work page doi:10.1145/2009916.2009934 2011

  26. [26]

    arXiv preprint arXiv:2506.05176 , year=

    Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models , author=. arXiv preprint arXiv:2506.05176 , year=

    Pith/arXiv arXiv

  27. [27]

    2020 , eprint=

    Passage Re-ranking with BERT , author=. 2020 , eprint=

    2020

  28. [28]

    Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages =

    Liu, Shichen and Xiao, Fei and Ou, Wenwu and Si, Luo , title =. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages =. 2017 , isbn =. doi:10.1145/3097983.3098011 , abstract =

    work page doi:10.1145/3097983.3098011 2017

  29. [29]

    The Science Behind Semantic Search: How

  30. [30]

    Retrieval-augmented generation for knowledge-intensive NLP tasks , year =

    Lewis, Patrick and Perez, Ethan and Piktus, Aleksandra and Petroni, Fabio and Karpukhin, Vladimir and Goyal, Naman and K\". Retrieval-augmented generation for knowledge-intensive NLP tasks , year =. Proceedings of the 34th International Conference on Neural Information Processing Systems , articleno =

  31. [31]

    2024 , note=

    Rerankers and Two-Stage Retrieval , author=. 2024 , note=

    2024

  32. [32]

    2024 , note=

    Amazon. 2024 , note=

    2024

  33. [33]

    2025 , eprint=

    Qwen3 Technical Report , author=. 2025 , eprint=

    2025

  34. [34]

    2024 , issue_date =

    Su, Jianlin and Ahmed, Murtadha and Lu, Yu and Pan, Shengfeng and Bo, Wen and Liu, Yunfeng , title =. 2024 , issue_date =. doi:10.1016/j.neucom.2023.127063 , journal =

    work page doi:10.1016/j.neucom.2023.127063 2024

  35. [35]

    https://huggingface.co/jinaai/jina-reranker-v2-base-multilingual

    jinaai/jina-reranker-v2-base-multilingual · Hugging Face , url = "https://huggingface.co/jinaai/jina-reranker-v2-base-multilingual", month =

  36. [36]

    2025 , note=

    Boost your Search and. 2025 , note=

    2025

  37. [37]

    URL https://aclanthology.org/2024.tacl-1.9/

    Liu, Nelson F. and Lin, Kevin and Hewitt, John and Paranjape, Ashwin and Bevilacqua, Michele and Petroni, Fabio and Liang, Percy. Lost in the Middle: How Language Models Use Long Contexts. Transactions of the Association for Computational Linguistics. 2024. doi:10.1162/tacl_a_00638

    work page doi:10.1162/tacl_a_00638 2024

  38. [38]

    First Conference on Language Modeling , year=

    How Easily do Irrelevant Inputs Skew the Responses of Large Language Models? , author=. First Conference on Language Modeling , year=

  39. [39]

    Introducing reranking to Pinecone Inference , author =

  40. [40]

    Is ChatGPT good at search? investigating large language models as re-ranking agents

    Sun, Weiwei and Yan, Lingyong and Ma, Xinyu and Wang, Shuaiqiang and Ren, Pengjie and Chen, Zhumin and Yin, Dawei and Ren, Zhaochun. Is C hat GPT Good at Search? Investigating Large Language Models as Re-Ranking Agents. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023. doi:10.18653/v1/2023.emnlp-main.923

    work page doi:10.18653/v1/2023.emnlp-main.923 2023

  41. [41]

    The Thirteenth International Conference on Learning Representations , year=

    Attention in Large Language Models Yields Efficient Zero-Shot Re-Rankers , author=. The Thirteenth International Conference on Learning Representations , year=

  42. [42]

    InFindings of the As- sociation for Computational Linguistics: ACL 2024, Lun-Wei Ku, Andre Martins, and Vivek Srikumar (Eds.)

    Chen, Jianlyu and Xiao, Shitao and Zhang, Peitian and Luo, Kun and Lian, Defu and Liu, Zheng. M 3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation. Findings of the Association for Computational Linguistics: ACL 2024. 2024. doi:10.18653/v1/2024.findings-acl.137

    work page doi:10.18653/v1/2024.findings-acl.137 2024

  43. [43]

    2023 , eprint=

    Making Large Language Models A Better Foundation For Dense Retrieval , author=. 2023 , eprint=

    2023

  44. [44]

    Is Anisotropy Truly Harmful? A Case Study on Text Clustering

    Ait-Saada, Mira and Nadif, Mohamed. Is Anisotropy Truly Harmful? A Case Study on Text Clustering. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). 2023. doi:10.18653/v1/2023.acl-short.103

    work page doi:10.18653/v1/2023.acl-short.103 2023

  45. [45]

    CoRR , volume=

    Shengyao Zhuang and Honglei Zhuang and Bevan Koopman and Guido Zuccon , title=. CoRR , volume=. 2023 , cdate=

    2023

  46. [46]

    The Shape of Learning: Anisotropy and Intrinsic Dimensions in Transformer-Based Models

    Razzhigaev, Anton and Mikhalchuk, Matvey and Goncharova, Elizaveta and Oseledets, Ivan and Dimitrov, Denis and Kuznetsov, Andrey. The Shape of Learning: Anisotropy and Intrinsic Dimensions in Transformer-Based Models. Findings of the Association for Computational Linguistics: EACL 2024. 2024

    2024

  47. [47]

    Anisotropy Is Inherent to Self-Attention in Transformers

    Godey, Nathan and Clergerie, \'E ric and Sagot, Beno \^i t. Anisotropy Is Inherent to Self-Attention in Transformers. Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers). 2024. doi:10.18653/v1/2024.eacl-long.3

    work page doi:10.18653/v1/2024.eacl-long.3 2024

  48. [48]

    Shrink the Longest: Improving Latent Space Isotropy with Simplicial Geometry

    Kudrjashov, Sergej and Karpik, Olesya and Klyshinsky, Eduard. Shrink the Longest: Improving Latent Space Isotropy with Simplicial Geometry. Analysis of Images, Social Networks and Texts. 2025

    2025

  49. [49]

    2024 , cdate=

    Xueguang Ma and Liang Wang and Nan Yang and Furu Wei and Jimmy Lin , title=. 2024 , cdate=

    2024

  50. [50]

    2026 , eprint=

    Rethinking the Reranker: Boundary-Aware Evidence Selection for Robust Retrieval-Augmented Generation , author=. 2026 , eprint=

    2026

  51. [51]

    CoRR , volume=

    Siyuan Meng and Junming Liu and Yirong Chen and Song Mao and Pinlong Cai and Guohang Yan and Botian Shi and Ding Wang , title=. CoRR , volume=. 2025 , month=

    2025

  52. [52]

    Forty-second International Conference on Machine Learning , year=

    Layer by Layer: Uncovering Hidden Representations in Language Models , author=. Forty-second International Conference on Machine Learning , year=

  53. [53]

    Constructing

    Ho, Xanh and Duong Nguyen, Anh-Khoa and Sugawara, Saku and Aizawa, Akiko. Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps. Proceedings of the 28th International Conference on Computational Linguistics. 2020. doi:10.18653/v1/2020.coling-main.580

    work page doi:10.18653/v1/2020.coling-main.580 2020

  54. [54]

    H otpot QA : A Dataset for Diverse, Explainable Multi-hop Question Answering

    Yang, Zhilin and Qi, Peng and Zhang, Saizheng and Bengio, Yoshua and Cohen, William and Salakhutdinov, Ruslan and Manning, Christopher D. H otpot QA : A Dataset for Diverse, Explainable Multi-hop Question Answering. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. 2018. doi:10.18653/v1/D18-1259

    work page doi:10.18653/v1/d18-1259 2018

  55. [55]

    ♫ M u S i Q ue: Multihop Questions via Single-hop Question Composition

    Trivedi, Harsh and Balasubramanian, Niranjan and Khot, Tushar and Sabharwal, Ashish. M u S i Q ue: Multihop Questions via Single-hop Question Composition. Transactions of the Association for Computational Linguistics. 2022. doi:10.1162/tacl_a_00475

    work page doi:10.1162/tacl_a_00475 2022

  56. [56]

    WWW'18 Open Challenge: Financial Opinion Mining and Question Answering , year =

    Maia, Macedo and Handschuh, Siegfried and Freitas, Andr\'. WWW'18 Open Challenge: Financial Opinion Mining and Question Answering , year =. Companion Proceedings of the The Web Conference 2018 , pages =. doi:10.1145/3184558.3192301 , abstract =

    work page doi:10.1145/3184558.3192301 2018

  57. [57]

    InCoCo@NIPS 2016 (Workshop at NIPS/NeurIPS 2016) , year=

    MS MARCO: A Human Generated MAchine Reading COmprehension Dataset , author=. InCoCo@NIPS 2016 (Workshop at NIPS/NeurIPS 2016) , year=

    2016

  58. [58]

    2018 , howpublished =

    Quora Question Pairs Dataset , author =. 2018 , howpublished =

    2018

  59. [59]

    FEVER: a large-scale dataset for Fact Extraction and VERification

    Thorne, James and Vlachos, Andreas and Christodoulopoulos, Christos and Mittal, Arpit. FEVER : a Large-scale Dataset for Fact Extraction and VER ification. Proceedings of the 2018 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018. doi:10.18653/v1/N18-1074

    work page internal anchor Pith review doi:10.18653/v1/n18-1074 2018

  60. [60]

    Proceedings of the 38th European Conference on Information Retrieval (ECIR) , year =

    Boteva, Vera and Gholipour, Demian and Sokolov, Artem and Riezler, Stefan , title =. Proceedings of the 38th European Conference on Information Retrieval (ECIR) , year =

  61. [61]

    2023 , eprint=

    RoFormer: Enhanced Transformer with Rotary Position Embedding , author=. 2023 , eprint=

    2023

  62. [62]

    Advances in Large Margin Classifiers , editor=

    Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods , author=. Advances in Large Margin Classifiers , editor=. 1999 , pages=

    1999