Pith. sign in

REVIEW 5 major objections 6 minor 36 references

Context Filtering with Reward Modeling in Question Answering

T0 review · 5 major / 6 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 Reward-model-based context filtering can cut retrieved passages to a small fraction of their length while keeping most question-answering accuracy, raising accuracy-per-token by up to 6.8 times.

desk verdict The efficiency claim is undercut by a likely out-of-distribution baseline for the reader, but the automatic preference-pair idea is worth a second look. read the letter →

arxiv 2412.11707 v1 pith:H2MEXLC6 submitted 2024-12-16 cs.CL

Sangryul Kim , James Thorne This is my paper · ORCID
classification cs.CL
keywords contextfilteringquestionansweringrewardmodelingdirectpreferenceoptimizationtokenefficiencyempersummarize-then-readdetrimentalretrieval
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

Context filtering in question answering usually means deciding which parts of a retrieved passage are worth keeping. The paper claims that this filtering can be learned with reward modeling instead of human labels: a summarization model is trained with Direct Preference Optimization on pairs where the 'chosen' summary was generated with the question, answer, and context all present, and the 'rejected' summaries were generated with one element missing. On SQuAD, Natural Questions, and TriviaQA, the resulting pipeline keeps most of the original accuracy while shortening the context to 12-20 percent of its original length. The central evidence is the new EM Per Token metric: dividing Exact Match by context length shows up to a 6.8-fold efficiency gain over using the full retrieved context. If true, this gives low-resource or token-budget settings a way to get more correct answers per token without expensive preference annotation.

What carries the argument

The central mechanism is the DPO preference-construction scheme built on the information tuple I=(Q,A,C). The authors define three prompt types: Type 1 feeds all three elements and produces the 'chosen' summary; Type 2 omits the answer; Type 3 omits the question, and summaries from these two types serve as 'rejected' outputs. DPO training then maximizes the margin between chosen and rejected summaries in reward space, and the resulting summarizer is used in a summarize-then-read pipeline. The other key object is the proposed EPT metric, $EPT(c, y^*, \hat{y}) = EM(y^*, \hat{y}) / |c|$, which is what the 6.8-fold claim is measured on.

What would settle it

Retrain the DPO summarizer after randomly flipping half of the chosen/rejected labels in the generated preference pairs; if EPT still improves by a comparable factor, the reward signal is not what drives the efficiency gain.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that answer-relevant filtering can be induced from the data itself: a summarizer trained to imitate complete-information summaries while being pushed away from incomplete-information summaries learns to keep the span the reader needs. The chosen and rejected outputs are generated from the same base model using three prompt types that combine or omit the question, answer, and context, and the DPO loss widens the reward gap between them. When the filtered contexts are fed to a trained reader, the model attains exact-match scores in the high 40s to high 60s on the three datasets, compared with high 50s to high 70s for the original contexts, but with roughly one-fifth to one-eighth the tokens, and the EPT values for the best DPO variant reach 2.58-3.42 versus 0.38-0.51 for the original contexts. The authors conclude that the efficiency gained from filtering outweighs the slight loss in accuracy.

Load-bearing premise

The load-bearing premise is that a summary generated from the full question-answer-context tuple is always the 'chosen' one and summaries missing either the question or the answer are always 'rejected', an assumption the authors concede may introduce unintended biases.

Editorial extensions

If this is right

  • A QA system with a fixed token budget can answer more questions correctly by filtering contexts with this method than by feeding full retrieved passages, making it attractive for API-based and on-device deployment.
  • The self-generated preference pairs remove the need for human preference annotation, so the training recipe can be rerun on any QA dataset where gold answers are available.
  • The summarize-then-read pipeline can be extended to multi-hop and long-context QA, settings the paper explicitly notes have not yet been tested.
  • The EPT metric gives a common scale for comparing context efficiency across models, so future filtering and compression work can be evaluated on accuracy per token rather than raw exact match.

Reading between the lines

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

  • Our inference: because the preference pairs are generated from the presence of the answer in the prompt, the method is likely to favor contexts where the answer appears lexically and may underperform in reasoning-heavy answering regimes; a reader-aware reward would be a natural correction.
  • Our inference: the EPT gain should be re-baselined against simply truncating contexts to the same lengths; if random-length truncation achieves comparable EPT, the value of the DPO filtering is in content selection rather than shortening alone.
  • Our inference: the same auto-generated preference scheme could be applied to other input-compression tasks such as retrieval-augmented generation with long documents, where the 'chosen' output is the full query-gold answer-document tuple and 'rejected' outputs are missing one component.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes a context-filtering approach for extractive question answering in which a FLAN-T5 summarizer is fine-tuned, first with SFT and then with DPO, to compress retrieved contexts before they are passed to a reader. DPO preference pairs are generated automatically from prompt variants that include or omit the question, answer, or context, without human annotation. The authors introduce an Exact Match per Token (EPT) metric and report that filtered contexts substantially improve EPT, including a claimed 6.8-fold improvement, while retaining most of the original EM accuracy. The paper concludes that the token-efficiency gains from filtering outweigh a slight accuracy loss.

Significance. If the findings were sound, the paper would offer a practical pipeline for lowering the inference cost of open-domain QA through learned context compression, and the EPT metric would be a useful complement to EM/F1 for comparing context efficiency. The release of code and data is also a positive factor. However, the evaluation protocol as presented does not support the headline efficiency claim: the baseline reader appears to be out-of-distribution, the EPT numbers in Table 1 do not obviously match Eq. (2), and the paper does not compare with existing context-compression baselines. The central contribution is therefore promising but not yet established.

major comments (5)
  1. [§3.4 and Table 1] The reader is trained only on filtered contexts ('we train the reader to generate answers using the filtered context'), yet the Origin row in Table 1 evaluates the same reader on original, long contexts that it never saw in training. The Origin baseline is therefore out-of-distribution for the reader, so the lower EM and EPT of Origin may reflect distribution shift rather than filtering quality. This directly undermines the Section 4 claim that efficiency gains outweigh accuracy loss. The authors should train a matched reader on original contexts and compare all conditions under the same training distribution, or otherwise control for this mismatch.
  2. [§3.5, Eq. (2), and Table 1] Eq. (2) defines EPT as EM(y*, yhat)/|c| per example, but the EPT values in Table 1 match dataset-level EM divided by average token length (e.g., 59.59/147.30 = 0.40 for NQr Origin). These are not the same quantity unless EM and token length are uncorrelated across examples. The 6.8-fold headline number is derived from this table-level computation, so the paper's primary quantitative claim is not backed by the stated metric. The authors should report the mean of per-example EPT and, ideally, confidence intervals.
  3. [§3.3 and Limitations] The DPO chosen/rejected pairs are generated by the model itself using Type 1 vs. Type 2/Type 3 prompts, with no human validation that the Type 1 summaries are actually preferred. Because the Type 1 prompt includes the answer, the chosen summaries contain the answer by construction, so DPO directly optimizes a proxy—answer inclusion—that is correlated with the downstream IRA and EM on filtered contexts. The authors themselves concede in the Limitations section that this data generation 'may introduce ... some unintended biases, leading to biased results and inconsistent outcomes across the datasets.' This concession should be taken as a load-bearing caveat: the reported DPO results cannot be interpreted as a validated filtering preference without additional control experiments or preference validation.
  4. [§4 and Table 1] The paper compares only against the original context and against SFT/DPO variants; it does not compare with existing context-filtering or compression baselines such as Wang et al. (2023) or Xu et al. (2023), both of which are cited in the text. Without such comparisons, the claim of 'significantly outperform[ing] the baseline' is not established relative to the state of the art. At minimum, the authors should include an extractive filtering baseline and a length-controlled baseline, and report variance or significance over multiple runs.
  5. [Abstract and §4] The abstract's claim of a 6.8-fold increase in EPT is presented as 'significantly outperform[ing] the baseline,' but the corresponding rows in Table 1 show substantial EM decreases (e.g., TQAr Origin 77.38 vs. DPO_O1,O3 62.29; SQuADr Origin 68.32 vs. DPO_O1,O3 55.40). Calling this a 'slight loss in accuracy' in Section 4 understates the trade-off. The authors should either separate efficiency claims from accuracy claims, or define a combined metric that explicitly makes the trade-off and justify why the efficiency gain dominates.
minor comments (6)
  1. [§3.3] The word 'Hense' should be 'Hence'.
  2. [§4] The word 'accuray' should be 'accuracy'.
  3. [§2.2] The phrase 'for a given given task prompt' contains a duplicated 'given'.
  4. [§3.5 and Eq. (2)] The notation in Eq. (2) is inconsistent with the surrounding text: the metric is called EPT but the equation reads 'EP T', and the variables y* and yhat are not defined until the next sentence.
  5. [Table 3 and Appendix C] The parenthetical numbers in the NQr and TQAr rows are not explained in the table caption; the text should clarify that these are the counts after the top-1 context filtering described in §3.1.
  6. [Figure 1] The example text in Figure 1 is rendered without spaces (e.g., 'ACountryBoyCanSurvive'), which makes the figure hard to read; please reformat it.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the reported EPT gains are arithmetic consequences of the paper's own evaluation metric, not a derivation that reduces to the training inputs.

full rationale

The paper's central efficiency claim is a measured comparison under the metric it defines, EPT = EM / |c| (Eq. 2), rather than a prediction derived from the method's assumptions. Dividing by context length will mechanically reward shorter contexts, and the 6.8x figure follows from the token-length reductions reported in Table 1, but this is the intended definition of the metric and not a hidden equivalence between an input and an output. The DPO training pairs are indeed constructed by the authors' assumption that Type 1 summaries (containing the answer) are 'chosen' and Type 2/3 summaries (missing information) are 'rejected', so the reward signal is aligned with answer inclusion; however, the main evaluation uses held-out exact-match answers from standard QA datasets, so the reported EM and EPT values are genuine measurements of downstream reader performance rather than quantities forced by the training label construction. The reader training on filtered contexts may make the Origin baseline out-of-distribution, and the authors' own Limitations section concedes that the self-generated preference data 'may introduce ... some unintended biases, leading to biased results and inconsistent outcomes across the datasets.' These are experimental validity and distribution-shift concerns, not circularity under the criteria here. Citations to the authors' prior work on detrimental retrieval are background motivation and are accompanied by independent external references, so no load-bearing claim reduces to a self-citation chain.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central claim rests on several domain assumptions about the validity of auto-generated preference pairs, the quality of the base model's summaries, and the meaningfulness of the EPT metric. No new entities are introduced, and no fitted numerical parameters are used besides standard training hyperparameters.

assumptions (5)
  • domain assumption DPO loss formulation (Eq. 1) from Rafailov et al. (2023) is valid for preference pairs.
    The paper builds its reward modeling directly on the DPO loss without re-deriving or validating its assumptions in the QA context.
  • domain assumption Type 1 summaries (with question, answer, context) are always preferred over Type 2 or Type 3 summaries for QA.
    This is the core premise of Section 3.3; the authors later admit in Limitations that this data generation may introduce bias.
  • domain assumption FLAN-T5-XL generated summaries are reliable training targets without human evaluation.
    The paper explicitly says it relies on the model's parameters without human evaluation, and the Limitations note introduces unintended biases.
  • domain assumption The EPT metric, EM divided by context length, is a meaningful measure of token efficiency.
    Equation 2 defines EPT as EM/|c|; the paper treats this as a valid efficiency indicator without justifying the linear trade-off.
  • domain assumption The reader model can be fairly evaluated on both original and summarized contexts despite being trained only on summarized contexts.
    Section 3.4 describes reader training on filtered contexts, but the Origin baseline is not clearly matched, leaving the comparison under-specified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Context Filtering with Reward Modeling in Question Answering." pith.science (2026). https://pith.science/paper/H2MEXLC6

@misc{pith2026241211707,
  author       = {Pith},
  title        = {Pith review of: Context Filtering with Reward Modeling in Question Answering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H2MEXLC6}},
  note         = {Machine review of arXiv:2412.11707}
}
read the original abstract

Question Answering (QA) in NLP is the task of finding answers to a query within a relevant context retrieved by a retrieval system. Yet, the mix of relevant and irrelevant information in these contexts can hinder performance enhancements in QA tasks. To address this, we introduce a context filtering approach that removes non-essential details, summarizing crucial content through Reward Modeling. This method emphasizes keeping vital data while omitting the extraneous during summarization model training. We offer a framework for developing efficient QA models by discerning useful information from dataset pairs, bypassing the need for costly human evaluation. Furthermore, we show that our approach can significantly outperform the baseline, as evidenced by a 6.8-fold increase in the EM Per Token (EPT) metric, which we propose as a measure of token efficiency, indicating a notable token-efficiency boost for low-resource settings.

Figures

Figures reproduced from arXiv: 2412.11707 by the authors.

Figure 1
Figure 1. For an effective QA task, we conduct con [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 10 canonical work pages

  1. [1]

    Kosuke Akimoto, Kunihiro Takeoka, and Masafumi Oyamada. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.784 Context quality matters in training fusion-in-decoder for extractive open-domain question answering . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 11711--11729, Singapore. Association for Computational Linguistics

    work page doi:10.18653/v1/2023.findings-emnlp.784 2023

  2. [2]

    Peters, and Arman Cohan

    Iz Beltagy, Matthew E. Peters, and Arman Cohan. 2020. https://arxiv.org/abs/2004.05150 Longformer: The long-document transformer . Preprint, arXiv:2004.05150

    arXiv 2020

  3. [3]

    Ralph Allan Bradley and Milton E. Terry. 1952. http://www.jstor.org/stable/2334029 Rank analysis of incomplete block designs: I. the method of paired comparisons . Biometrika, 39(3/4):324--345

    arXiv 1952

  4. [4]

    Avi Caciularu, Ido Dagan, Jacob Goldberger, and Arman Cohan. 2022. https://doi.org/10.18653/v1/2022.naacl-main.207 Long context question answering via supervised contrastive learning . In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 2872--2879, Seattle...

    work page doi:10.18653/v1/2022.naacl-main.207 2022

  5. [5]

    Andong Chen, Yuan Sun, Xiaobing Zhao, Rosella Galindo Esparza, Kehai Chen, Yang Xiang, Tiejun Zhao, and Min Zhang. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.699 Improving low-resource question answering by augmenting question information . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 10413--10420, Singapore. ...

    work page doi:10.18653/v1/2023.findings-emnlp.699 2023

  6. [6]

    Danqi Chen, Adam Fisch, Jason Weston, and Antoine Bordes. 2017. https://doi.org/10.18653/v1/P17-1171 Reading W ikipedia to answer open-domain questions . In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1870--1879, Vancouver, Canada. Association for Computational Linguistics

    doi:10.18653/v1/p17-1171 2017

  7. [7]

    Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al. 2022. Scaling instruction-finetuned language models. arXiv preprint arXiv:2210.11416

    arXiv 2022

  8. [8]

    Nicola De Cao , Gautier Izacard, Sebastian Riedel, and Fabio Petroni. 2021. https://openreview.net/forum?id=5k8F6UU39V Autoregressive entity retrieval . In 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021 . OpenReview.net

    work page 2021

Show all 36 references
  1. [9]

    Angela Fan, Yacine Jernite, Ethan Perez, David Grangier, Jason Weston, and Michael Auli. 2019. https://doi.org/10.18653/v1/P19-1346 ELI 5: Long form question answering . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3558--356...

    2019 doi

  2. [10]

    Yair Feldman and Ran El-Yaniv. 2019. https://doi.org/10.18653/v1/P19-1222 Multi-hop paragraph retrieval for open-domain question answering . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 2296--2309, Florence, Italy. Associati...

    2019 doi

  3. [11]

    Chloe Ching-Yun Hsu, Celestine Mendler-D \"u nner, and Moritz Hardt. 2020. Revisiting design choices in proximal policy optimization. arXiv preprint arXiv:2009.10897

    2020 arXiv

  4. [12]

    Naoya Inoue, Harsh Trivedi, Steven Sinha, Niranjan Balasubramanian, and Kentaro Inui. 2021. https://doi.org/10.18653/v1/2021.emnlp-main.490 Summarize-then-answer: Generating concise explanations for multi-hop reading comprehension . In Proceedings of the 2021 Conference on Emp...

    2021 doi

  5. [13]

    Gautier Izacard and Edouard Grave. 2021. https://doi.org/10.18653/v1/2021.eacl-main.74 Leveraging passage retrieval with generative models for open domain question answering . In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Li...

    2021 doi

  6. [14]

    Mandar Joshi, Eunsol Choi, Daniel Weld, and Luke Zettlemoyer. 2017. https://doi.org/10.18653/v1/P17-1147 T rivia QA : A large scale distantly supervised challenge dataset for reading comprehension . In Proceedings of the 55th Annual Meeting of the Association for Computational...

    2017 doi

  7. [15]

    Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. https://doi.org/10.18653/v1/2020.emnlp-main.550 Dense passage retrieval for open-domain question answering . In Proceedings of the 2020 Conference on Empiric...

    2020 doi

  8. [16]

    Dai, Jakob Uszkoreit, Quoc Le, and Slav Petrov

    Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, Kristina Toutanova, Llion Jones, Matthew Kelcey, Ming-Wei Chang, Andrew M. Dai, Jakob Uszkoreit, Quoc Le, and Slav...

    2019 doi

  9. [17]

    Yejoon Lee, Philhoon Oh, and James Thorne. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.616 Knowledge corpus error in question answering . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9183--9197, Singapore. Association for Computatio...

    2023 doi

  10. [18]

    Arnold, and Bing Xiang

    Feng Nan, Cicero Nogueira dos Santos, Henghui Zhu, Patrick Ng, Kathleen McKeown, Ramesh Nallapati, Dejiao Zhang, Zhiguo Wang, Andrew O. Arnold, and Bing Xiang. 2021. https://doi.org/10.18653/v1/2021.acl-long.536 Improving factual consistency of abstractive summarization via qu...

    2021 doi

  11. [19]

    Philhoon Oh and James Thorne. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.776 Detrimental contexts in open-domain question answering . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 11589--11605, Singapore. Association for Computation...

    2023 doi

  12. [20]

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 3...

    2022

  13. [21]

    Fabio Petroni, Aleksandra Piktus, Angela Fan, Patrick Lewis, Majid Yazdani, Nicola De Cao, James Thorne, Yacine Jernite, Vladimir Karpukhin, Jean Maillard, Vassilis Plachouras, Tim Rockt \"a schel, and Sebastian Riedel. 2021. https://doi.org/10.18653/v1/2021.naacl-main.200 KIL...

    2021 doi

  14. [22]

    Rafael Rafailov, Joey Hejna, Ryan Park, and Chelsea Finn. 2024. https://openreview.net/forum?id=kEVcNxtqXk From \ r\ to \ q *\ : Your language model is secretly a q-function . In First Conference on Language Modeling

    2024

  15. [23]

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Stefano Ermon, Christopher D Manning, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. arXiv preprint arXiv:2305.18290

    2023 arXiv

  16. [24]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. http://jmlr.org/papers/v21/20-074.html Exploring the limits of transfer learning with a unified text-to-text transformer . Journal of Machine Lea...

    2020

  17. [25]

    Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. https://doi.org/10.18653/v1/D16-1264 SQ u AD : 100,000+ questions for machine comprehension of text . In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2383-...

    2016 doi

  18. [26]

    Artsiom Sauchuk, James Thorne, Alon Halevy, Nicola Tonellotto, and Fabrizio Silvestri. 2022. On the role of relevance in natural language processing tasks. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages...

    2022

  19. [27]

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347

    2017 arXiv

  20. [28]

    Freda Shi, Xinyun Chen, Kanishka Misra, Nathan Scales, David Dohan, Ed Chi, Nathanael Sch\" a rli, and Denny Zhou. 2023. Large language models can be easily distracted by irrelevant context. In Proceedings of the 40th International Conference on Machine Learning, ICML'23. JMLR.org

    2023

  21. [29]

    Leandro von Werra, Younes Belkada, Lewis Tunstall, Edward Beeching, Tristan Thrush, Nathan Lambert, and Shengyi Huang. 2020. Trl: Transformer reinforcement learning. https://github.com/huggingface/trl

    2020

  22. [30]

    Zhiruo Wang, Jun Araki, Zhengbao Jiang, Md Rizwan Parvez, and Graham Neubig. 2023. Learning to filter context for retrieval-augmented generation. arXiv preprint arXiv:2311.08377

    2023 arXiv

  23. [31]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mari...

    2020 doi

  24. [32]

    Fangyuan Xu, Weijia Shi, and Eunsol Choi. 2023. https://arxiv.org/abs/2310.04408 Recomp: Improving retrieval-augmented lms with compression and selective augmentation . Preprint, arXiv:2310.04408

    2023 arXiv

  25. [33]

    Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. https://doi.org/10.18653/v1/D18-1259 H otpot QA : A dataset for diverse, explainable multi-hop question answering . In Proceedings of the 2018 Conference...

    2018 doi

  26. [34]

    Wenhao Yu, Dan Iter, Shuohang Wang, Yichong Xu, Mingxuan Ju, Soumya Sanyal, Chenguang Zhu, Michael Zeng, and Meng Jiang. 2022. Generate rather than retrieve: Large language models are strong context generators. arXiv preprint arXiv:2209.10063

    2022 arXiv

  27. [35]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  28. [36]

    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.