Pith. sign in

REVIEW 4 major objections 4 minor 43 references

M-BRe: Discovering Training Samples for Relation Extraction from Unlabeled Texts with Large Language Models

T0 review · 4 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read M-BRe can construct relation-extraction training samples from unlabeled text at about half the LLM running time of per-relation binary classification while matching its accuracy.

desk verdict Practical idea, real downstream gains, but unlabeled-corpus provenance is unspecified and the framework's own evaluation is thin. read the letter →

arxiv 2509.07730 v2 pith:KBFV2QMK submitted 2025-09-09 cs.CL

Zexuan Li , Hongliang Dai , Piji Li This is my paper
classification cs.CL
keywords relationextractionlargelanguagemodelstrainingdatadiscoveryunlabeledtextfew-shotlearningmulti-classclassificationbinarylabeldecision
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

Manually labeling relation-extraction training data is costly because sentences containing a target relation are rare in raw text. The paper's M-BRe framework uses an LLM to find such sentences automatically: it groups similar relation categories, asks the LLM to classify within each group, then verifies each candidate label with a binary question. The paper claims this two-stage design matches the accuracy of running binary classification for every relation while taking less than half the time. It further reports that mixing the discovered samples with a few manually labeled examples significantly improves conventional relation-extraction models, and that the discovered samples alone rival small manual sets in 2- and 4-shot settings. If these claims hold, relation-extraction datasets can be bootstrapped from unlabeled text at a fraction of the annotation cost.

What carries the argument

Three modules carry the argument. Relation Grouping embeds short relation explanations, builds a cosine-similarity matrix, seeds the two most dissimilar relations, and greedily assigns the rest to minimize maximum within-group similarity, producing floor(N/6) groups. Relation Extraction builds one multi-class prompt per group, plus binary verification prompts for each candidate relation, both using in-context demonstrations. Label Decision computes confidence as the mean maximum softmax probability over generated tokens and accepts any candidate whose confidence passes 1 minus 10^-2, handling the no-yes, single-yes, and multi-yes cases. The grouping is the load-bearing mechanism: it converts

What would settle it

Check whether any of the 4,401 unlabeled sentences, or near-duplicates of them, appear in the TACRED, TACRED-Revisit, Re-TACRED, or SemEval test splits; if a material fraction of the generated training samples match test sentences, the reported gains from mixing generated and manual samples are inflated by test-set leakage. A cleaner test: run M-BRe on unlabeled text from a domain different from the benchmark and see whether trained models still beat manual-only few-shot baselines.

Watch

Extended reading notes

Core claim

The central discovery is that a grouped hybrid of multi-class and binary classification lets LLMs mine usable relation-extraction training instances from open-domain unlabeled sentences. Given N relation categories, M-BRe partitions them into about N/6 groups, seeding with the two least similar relations and greedily assigning the rest to keep within-group similarity low. For each sentence-entity pair, one multi-class prompt per group produces candidate relations, and a binary prompt checks each candidate. A confidence score, the average maximum softmax probability over generated tokens, decides between 'no relation' and one or more accepted labels. On TACRED, TACRED-Revisit, Re-TACRED, and

Load-bearing premise

The 4,401 unlabeled sentences used to construct training samples are assumed to be disjoint from the test sets and free of near-duplicate leakage; the paper reports only the count, never the source, of these sentences.

Editorial extensions

If this is right

  • Unlabeled text becomes a practical source of RE training data: a practitioner with a corpus and a relation definition list can produce thousands of labeled instances without manual annotation beyond prompt design.
  • The speed-accuracy trade-off hinges on group count: at about floor(N/6) groups, quality matches exhaustive binary checks at less than half the time, and quality degrades if groups are pushed further.
  • Discovered samples are most valuable in the few-shot regime: mixing them with 2, 4, or 8 manual examples per relation beats manual-only training, but their marginal benefit fades as manual data grows.
  • A two-stage SFT schedule, discovered samples first then manual labels, slightly beats manual-only SFT for LLM multi-class RE, suggesting generated samples serve as a cold start.
  • The same pipeline transfers to fine-grained named entity recognition, where M-BRe raises recall over multi-class LLM prompting at moderate time cost.

Reading between the lines

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

  • The grouped-verification design should generalize beyond relation extraction to other many-class classification tasks where an LLM can propose candidates quickly but struggles to weigh all options at once; a testable variant would set group size from a confusion matrix rather than a fixed N/6 heuristic.
  • A cleaner test of the generated samples' value would compare M-BRe against random unlabeled sentences and against distant supervision at the same annotation budget, holding the unlabeled corpus fixed.
  • The long-tail issue the paper acknowledges suggests active selection over the unlabeled corpus: query sentences whose predicted relations are underrepresented, rather than sampling entity pairs uniformly.
  • The paper reports only the count of unlabeled sentences, not their provenance; users should verify that a target test corpus is disjoint from the unlabeled source before trusting reported gains.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes M-BRe, a framework that uses LLMs to discover relation-extraction training samples from unlabeled text. Relation types are first grouped by embedding similarity; the LLM performs a smaller multi-class classification within each group and then verifies each positive prediction with a binary prompt; a confidence-based decision module handles single and multiple positive labels. Experiments on TACRED, TACRED-Revisit, Re-TACRED, and SemEval with three Qwen models show that mixing the discovered samples with few-shot manual labels improves KnowPrompt and RetrievalRE, and that generated samples can serve as a cold-start for LLM supervised fine-tuning. The framework is also reported to be much faster than exhaustive binary classification.

Significance. The paper addresses a real cost bottleneck in relation extraction: finding training sentences for target relations in unlabeled text. The grouping idea is sensible and potentially practical, and the experimental scope is broad: three LLMs, four benchmarks, ablations on grouping, threshold sensitivity, and a downstream SFT study. The code is promised publicly, and the Limitations section candidly notes the long-tail and LLM-dependence issues. However, the central empirical claim is not yet fully established because the unlabeled corpus is not described, the reported comparisons lack statistical robustness, and key hyperparameters are chosen on the evaluation sets. These are fixable within the manuscript's scope, so the work merits revision rather than rejection.

major comments (4)
  1. [Appendix A.3; Tables 2 and 6] The only information about the 4,401 unlabeled sentences is their count. Their source, selection procedure, and relationship to the standard splits are not reported. If these sentences overlap or are near-duplicates of the TACRED/TACRED-Revisit/Re-TACRED/SemEval test sets, the gains in Tables 2 and 6 could reflect train/test leakage rather than the framework's discovery ability. The Limitations section does not address this. Please report the origin of the unlabeled corpus, how the 4,401 sentences were sampled, and an overlap/near-duplicate analysis against train/dev/test splits; re-run any affected experiments after removing overlapping items.
  2. [Section 4.2; Tables 2, 3, and 6] The direct framework evaluation uses only 207 TACRED and 91 SemEval samples, and no repeated seeds, error bars, or significance tests are reported anywhere in the paper. The downstream comparisons in Tables 2, 3, and 6 appear to be single runs. Given that LLM sampling temperature is 0.6 and few-shot fine-tuning is stochastic, statements such as 'significantly improves' and 'consistently outperform' are not supported. Please provide multiple seeds, standard deviations, and significance tests at least for the main comparisons in Tables 2 and 6 and for Figure 1.
  3. [Section 3.3; Table 1; Section 5.3] The confidence threshold θ=10^-2 and the group count P=⌊N/6⌋ are selected based on experiments conducted on the same datasets used for final reporting (Table 1, Figure 4, Table 13). This creates a selection bias: the reported numbers may be optimistically tuned. Please evaluate these hyperparameters on a held-out split or with a pre-registered rule, or otherwise show that the main conclusions are invariant over the explored grid.
  4. [Appendix C, Algorithm 1] The pseudocode as written does not produce a valid grouping. In the while loop, lines 7–9 compute a best group g* for each unassigned relation r_u, but r_u is never assigned to any group; line 10 then refers to an undefined r*. This makes the relation grouping algorithm non-reproducible from the paper alone. Please correct the pseudocode or explicitly point to the implementation.
minor comments (4)
  1. [Section 5.4; Table 6] The route labels 'Manual→Generated' and 'Generated→Manual' appear reversed relative to their arrow order. The text says Manual→Generated means generated-then-manual, which is the opposite of the natural reading. Since the cold-start conclusion depends on training order, please rename the columns (e.g., 'Gen→Manual' and 'Manual→Gen') consistently with the bullet list.
  2. [Section 4.2] The correctness criterion 'as long as the predicted relation list includes the ground truth' rewards multi-label outputs for listing many relations. Special_Avg_F1 mitigates this, but the headline Figure 1 comparison would be more informative if precision and recall of the final decision were also reported separately for M-BRe and the binary baseline, especially for the Multi-Yes cases.
  3. [Throughout] There are several typos and inconsistencies: 'B-MRe' appears instead of 'M-BRe' in the Limitations section and Appendix A.2; 'Tabel' in Table 1's caption; 'Precison_i' in the Special_Avg_F1 definition; and inconsistent capitalization of 'Tacred' vs. 'TACRED' in Appendix tables.
  4. [Section 3.3] The confidence formula is under-specified: 'max(softmax(logits_i))' should state whether the maximum is taken over the vocabulary for each generated token and how logits for different tokens are aggregated. Please clarify.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the framework is evaluated empirically with disclosed hyperparameter tuning, and no prediction reduces to its input by construction.

full rationale

The paper describes an empirical pipeline (Relation Grouping, Relation Extraction, Label Decision) and validates it by direct evaluation on sample test sentences and by downstream training of KnowPrompt/RetrievalRE and LLM SFT. There is no derivation chain whose conclusion is equivalent to its premises. The confidence threshold (epsilon=1e-2) and group count (floor(N/6)) are chosen using preliminary experiments on the same benchmarks, and Section 5.3/Table 1 disclose the selection and show robustness to these choices; this is hyperparameter tuning, not a fitted quantity being renamed as a prediction. The only self-citation (Li et al. 2025) appears in related-work context and is not load-bearing. The unspecified provenance of the 4401 unlabeled sentences (Appendix A.3) is a potential leakage risk, but the paper does not assert a provenance that would make the results definitionally circular, and no evidence is provided that the unlabeled sentences overlap the test sets. No imported uniqueness theorems or ansatz-by-citation are present. Hence no circular step can be exhibited under the required standard.

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

The framework introduces no new theoretical entities or forces. It depends on standard machine learning assumptions: LLM reliability, embedding-based similarity as a proxy for relation confusability, and distributional match between unlabeled text and target benchmarks. The two main tuned hyperparameters (group size k and confidence threshold theta) are selected on the target datasets.

free parameters (4)
  • Group size k = k = floor(N/6)
    Chosen by hand and empirically validated in Section 5.3. Performance peaks at this value on the same benchmarks, so it is a tuned hyperparameter.
  • Confidence threshold theta = 0.01
    Set in Section 3.3 based on preliminary experiments. Table 1 shows the optimal value varies by model and dataset; theta=0.01 is not always best (e.g., Qwen2.5-7B TACRED: 53.14 vs 55.07 at theta=0.001).
  • LLM sampling temperature = 0.6
    Set in Section A.2 without ablation.
  • Number of demonstrations in binary prompt = 3 correct, 4 incorrect
    Fixed in Section 3.2 without ablation.
assumptions (4)
  • domain assumption LLM annotations are accurate enough for training data construction
    The entire framework assumes LLM outputs (multi-class and binary) reflect true relations; only a small test subset is used for direct validation.
  • domain assumption TF-IDF vectorization of relation explanations captures semantic similarity
    Conditional on this, grouping by cosine similarity separates confusable relations; Section 3.1 and Algorithm 1.
  • domain assumption Unlabeled sentences are drawn from a distribution matching the target RE task
    The paper never specifies the corpus; the usefulness of generated samples depends on distributional match with the evaluation data.
  • domain assumption The 'predicted list includes ground truth' criterion is a valid proxy for extraction quality
    Adopted in Section 4.2; it is lenient and inflates recall.

how reviews work

0 comments
Cite this review

Pith. "Pith review of M-BRe: Discovering Training Samples for Relation Extraction from Unlabeled Texts with Large Language Models." pith.science (2026). https://pith.science/paper/KBFV2QMK

@misc{pith2026250907730,
  author       = {Pith},
  title        = {Pith review of: M-BRe: Discovering Training Samples for Relation Extraction from Unlabeled Texts with Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KBFV2QMK}},
  note         = {Machine review of arXiv:2509.07730}
}
read the original abstract

For Relation Extraction (RE), the manual annotation of training data may be prohibitively expensive, since the sentences that contain the target relations in texts can be very scarce and difficult to find. It is therefore beneficial to develop an efficient method that can automatically extract training instances from unlabeled texts for training RE models. Recently, large language models (LLMs) have been adopted in various natural language processing tasks, with RE also benefiting from their advances. However, when leveraging LLMs for RE with predefined relation categories, two key challenges arise. First, in a multi-class classification setting, LLMs often struggle to comprehensively capture the semantics of every relation, leading to suboptimal results. Second, although employing binary classification for each relation individually can mitigate this issue, it introduces significant computational overhead, resulting in impractical time complexity for real-world applications. Therefore, this paper proposes a framework called M-BRe to extract training instances from unlabeled texts for RE. It utilizes three modules to combine the advantages of both of the above classification approaches: Relation Grouping, Relation Extraction, and Label Decision. Extensive experiments confirm its superior capability in discovering high-quality training samples from unlabeled texts for RE.

Figures

Figures reproduced from arXiv: 2509.07730 by the authors.

Figure 1
Figure 1. Evaluation metrics across three different Rela [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Workflow of the M-BRe Framework, which consists of three modules: Relation Grouping Module, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Construction of Multi-Prompt and Binary-Prompt, which consists of three component: Task Description [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Micro F1 (%) and Special_Avg_F1 (%) of different number of groupings on each LLM. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Comparative evaluation of the two methods’ [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: A case study of relation extraction on sentence-level unlabeled text with the M-BRe framework, where [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

43 extracted references · 23 canonical work pages

  1. [1]

    Christoph Alt, Aleksandra Gabryszak, and Leonhard Hennig. 2020. https://doi.org/10.18653/V1/2020.ACL-MAIN.142 TACRED revisited: A thorough evaluation of the TACRED relation extraction task . In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020, Online, July 5-10, 2020 , pages 1558--1569. Association for Comp...

    doi:10.18653/v1/2020.acl-main.142 2020

  2. [2]

    Batista, Bruno Martins, and M \' a rio J

    David S. Batista, Bruno Martins, and M \' a rio J. Silva. 2015. https://doi.org/10.18653/V1/D15-1056 Semi-supervised bootstrapping of relationship extractors with distributional semantics . In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, EMNLP 2015, Lisbon, Portugal, September 17-21, 2015 , pages 499--504. The As...

    work page doi:10.18653/v1/d15-1056 2015

  3. [3]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert - Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litw...

    2020

  4. [4]

    Xiang Chen, Lei Li, Ningyu Zhang, Chuanqi Tan, Fei Huang, Luo Si, and Huajun Chen. 2022 a . https://doi.org/10.1145/3477495.3531746 Relation extraction as open-book examination: Retrieval-enhanced prompt tuning . In SIGIR '22: The 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, Madrid, Spain, July 11 - 15, 202...

    arXiv 2022

  5. [5]

    Xiang Chen, Ningyu Zhang, Xin Xie, Shumin Deng, Yunzhi Yao, Chuanqi Tan, Fei Huang, Luo Si, and Huajun Chen. 2022 b . https://doi.org/10.1145/3485447.3511998 Knowprompt: Knowledge-aware prompt-tuning with synergistic optimization for relation extraction . In WWW '22: The ACM Web Conference 2022, Virtual Event, Lyon, France, April 25 - 29, 2022 , pages 277...

    arXiv 2022

  6. [6]

    Yew Ken Chia, Lidong Bing, Soujanya Poria, and Luo Si. 2022. https://doi.org/10.18653/V1/2022.FINDINGS-ACL.5 Relationprompt: Leveraging prompts to generate synthetic data for zero-shot relation triplet extraction . In Findings of the Association for Computational Linguistics: ACL 2022, Dublin, Ireland, May 22-27, 2022 , pages 45--57. Association for Compu...

    doi:10.18653/v1/2022.findings-acl.5 2022

  7. [7]

    Jacob Devlin, Ming - Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://doi.org/10.18653/V1/N19-1423 BERT: pre-training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, ...

    doi:10.18653/v1/n19-1423 2019

  8. [8]

    Jiahui Gao, Renjie Pi, Yong Lin, Hang Xu, Jiacheng Ye, Zhiyong Wu, Weizhong Zhang, Xiaodan Liang, Zhenguo Li, and Lingpeng Kong. 2023. https://openreview.net/forum?id=h5OpjGd\_lo6 Self-guided noise-free data generation for efficient zero-shot learning . In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-...

    work page 2023

Show all 43 references
  1. [9]

    Tianyu Gao, Xu Han, Ruobing Xie, Zhiyuan Liu, Fen Lin, Leyu Lin, and Maosong Sun. 2020. https://doi.org/10.1609/AAAI.V34I05.6281 Neural snowball for few-shot relation learning . In The Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2020, The Thirty-Second Innov...

    2020 doi

  2. [10]

    Zhijiang Guo, Guoshun Nan, Wei Lu, and Shay B. Cohen. 2020. https://doi.org/10.24963/IJCAI.2020/505 Learning latent forests for medical relation extraction . In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI 2020 , pages 3651--...

    2020 doi

  3. [11]

    Zhijiang Guo, Yan Zhang, and Wei Lu. 2019. https://doi.org/10.18653/V1/P19-1024 Attention guided graph convolutional networks for relation extraction . In Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Florence, Italy, July 28- A...

    2019 doi

  4. [12]

    Thomas Hartvigsen, Saadia Gabriel, Hamid Palangi, Maarten Sap, Dipankar Ray, and Ece Kamar. 2022. https://doi.org/10.18653/V1/2022.ACL-LONG.234 Toxigen: A large-scale machine-generated dataset for adversarial and implicit hate speech detection . In Proceedings of the 60th Annu...

    2022 doi

  5. [13]

    Iris Hendrickx, Su Nam Kim, Zornitsa Kozareva, Preslav Nakov, Diarmuid \' O S \' e aghdha, Sebastian Pad \' o , Marco Pennacchiotti, Lorenza Romano, and Stan Szpakowicz. 2010. https://aclanthology.org/S10-1006/ Semeval-2010 task 8: Multi-way classification of semantic relation...

    2010

  6. [14]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen - Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen - Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. https://openreview.net/forum?id=nZeVKeeFYf9 Lora: Low-rank adaptation of large language models . In The Tenth International Conference on Learning Representat...

    2022

  7. [15]

    Weipeng Huang, Xingyi Cheng, Taifeng Wang, and Wei Chu. 2019. https://doi.org/10.1007/978-3-030-32236-6\_65 Bert-based multi-head selection for joint entity-relation extraction . In Natural Language Processing and Chinese Computing - 8th CCF International Conference, NLPCC 201...

    2019 doi

  8. [16]

    Guozheng Li, Peng Wang, and Wenjun Ke. 2023. https://doi.org/10.18653/V1/2023.FINDINGS-EMNLP.459 Revisiting large language models as zero-shot relation extractors . In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023 , pages...

    2023 doi

  9. [17]

    Zexuan Li, Hongliang Dai, and Piji Li. 2025. https://aclanthology.org/2025.acl-long.35/ Generating diverse training samples for relation extraction with large language models . In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1...

    2025

  10. [18]

    Yang Liu, Jinpeng Hu, Xiang Wan, and Tsung - Hui Chang. 2022. https://doi.org/10.18653/V1/2022.FINDINGS-NAACL.139 Learn from relation information: Towards prototype representation rectification for few-shot relation extraction . In Findings of the Association for Computational...

    2022 doi

  11. [19]

    Ye Liu, Kai Zhang, Aoran Gan, Linan Yue, Feng Hu, Qi Liu, and Enhong Chen. 2024. https://doi.org/10.1007/978-981-97-5569-1\_22 Empowering few-shot relation extraction with the integration of traditional RE methods and large language models . In Database Systems for Advanced Ap...

    2024 doi

  12. [20]

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. http://arxiv.org/abs/1907.11692 Roberta: A robustly optimized BERT pretraining approach . CoRR, abs/1907.11692

    2019 arXiv

  13. [21]

    Lin Long, Rui Wang, Ruixuan Xiao, Junbo Zhao, Xiao Ding, Gang Chen, and Haobo Wang. 2024. https://doi.org/10.18653/V1/2024.FINDINGS-ACL.658 On llms-driven synthetic data generation, curation, and evaluation: A survey . In Findings of the Association for Computational Linguisti...

    2024 doi

  14. [22]

    Yu Meng, Jiaxin Huang, Yu Zhang, and Jiawei Han. 2022. http://papers.nips.cc/paper\_files/paper/2022/hash/0346c148ba1c21c6b4780a961ea141dc-Abstract-Conference.html Generating training data with language models: Towards zero-shot language understanding . In Advances in Neural I...

    2022

  15. [23]

    Makoto Miwa and Mohit Bansal. 2016. https://doi.org/10.18653/V1/P16-1105 End-to-end relation extraction using lstms on sequences and tree structures . In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, ACL 2016, August 7-12, 2016, Berli...

    2016 doi

  16. [24]

    Mac \^ e do, Cleber Zanchettin, and Luciano Barbosa

    Johny Moreira, Chaina Oliveira, David L. Mac \^ e do, Cleber Zanchettin, and Luciano Barbosa. 2020. https://doi.org/10.1109/IJCNN48605.2020.9206648 Distantly-supervised neural relation extraction with side information using BERT . In 2020 International Joint Conference on Neur...

    2020

  17. [25]

    Giovanni Paolini, Ben Athiwaratkun, Jason Krone, Jie Ma, Alessandro Achille, Rishita Anubhai, C \' cero Nogueira dos Santos, Bing Xiang, and Stefano Soatto. 2021. https://openreview.net/forum?id=US-TP-xnXI Structured prediction as translation between augmented natural language...

    2021

  18. [26]

    Meng Qu, Tianyu Gao, Louis - Pascal A. C. Xhonneux, and Jian Tang. 2020. http://proceedings.mlr.press/v119/qu20a.html Few-shot relation extraction via bayesian meta-learning on relation graphs . In Proceedings of the 37th International Conference on Machine Learning, ICML 2020...

    2020

  19. [27]

    Laradji, Parmida Atighehchian, David V \' a zquez, and Dzmitry Bahdanau

    Gaurav Sahu, Pau Rodr \' guez, Issam H. Laradji, Parmida Atighehchian, David V \' a zquez, and Dzmitry Bahdanau. 2022. https://doi.org/10.18653/V1/2022.NLP4CONVAI-1.5 Data augmentation for intent classification with off-the-shelf large language models . In Proceedings of the 4...

    2022 doi

  20. [28]

    Srihari and Wei Li

    Rohini K. Srihari and Wei Li. 1999. http://trec.nist.gov/pubs/trec8/papers/cymfony.pdf Information extraction supported question answering . In Proceedings of The Eighth Text REtrieval Conference, TREC 1999, Gaithersburg, Maryland, USA, November 17-19, 1999 , volume 500-246 of...

    1999

  21. [29]

    George Stoica, Emmanouil Antonios Platanios, and Barnab \' a s P \' o czos. 2021. https://doi.org/10.1609/AAAI.V35I15.17631 Re-tacred: Addressing shortcomings of the TACRED dataset . In Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Thirty-Third Conference...

    2021 doi

  22. [30]

    Zhen Wan, Fei Cheng, Zhuoyuan Mao, Qianying Liu, Haiyue Song, Jiwei Li, and Sadao Kurohashi. 2023. https://doi.org/10.18653/V1/2023.EMNLP-MAIN.214 GPT-RE: in-context learning for relation extraction using large language models . In Proceedings of the 2023 Conference on Empiric...

    2023 doi

  23. [31]

    Chih-Hsuan Wei, Yifan Peng, Robert Leaman, Allan Peter Davis, Carolyn J Mattingly, Jiao Li, Thomas C Wiegers, and Zhiyong Lu. 2016. Assessing the state of the art in biomedical relation extraction: overview of the biocreative v chemical-disease relation (cdr) task. Database, 2...

    2016

  24. [32]

    Xin Xu, Yuqi Zhu, Xiaohan Wang, and Ningyu Zhang. 2023. https://doi.org/10.18653/V1/2023.SUSTAINLP-1.13 How to unleash the power of large language models for few-shot relation extraction? In Proceedings of The Fourth Workshop on Simple and Efficient Natural Language Processing...

    2023 doi

  25. [33]

    An Yang, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoyan Huang, Jiandong Jiang, Jianhong Tu, Jianwei Zhang, Jingren Zhou, Junyang Lin, Kai Dang, Kexin Yang, Le Yu, Mei Li, Minmin Sun, Qin Zhu, Rui Men, Tao He, Weijia Xu, Wenbiao Yin, Wenyuan Yu, Xiafei Qiu, Xingzhang R...

    2025 doi arXiv

  26. [34]

    Jiacheng Ye, Jiahui Gao, Qintong Li, Hang Xu, Jiangtao Feng, Zhiyong Wu, Tao Yu, and Lingpeng Kong. 2022. https://doi.org/10.18653/V1/2022.EMNLP-MAIN.801 Zerogen: Efficient zero-shot learning via dataset generation . In Proceedings of the 2022 Conference on Empirical Methods i...

    2022 doi

  27. [35]

    Daojian Zeng, Kang Liu, Yubo Chen, and Jun Zhao. 2015. https://doi.org/10.18653/V1/D15-1203 Distant supervision for relation extraction via piecewise convolutional neural networks . In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, EMNL...

    2015 doi

  28. [36]

    Kai Zhang, Bernal Jimenez Gutierrez, and Yu Su. 2023. https://doi.org/10.18653/V1/2023.FINDINGS-ACL.50 Aligning instruction tasks unlocks large language models as zero-shot relation extractors . In Findings of the Association for Computational Linguistics: ACL 2023, Toronto, C...

    2023 doi

  29. [37]

    Yuhao Zhang, Peng Qi, and Christopher D. Manning. 2018. https://doi.org/10.18653/V1/D18-1244 Graph convolution over pruned dependency trees improves relation extraction . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgi...

    2018 doi

  30. [38]

    Yuhao Zhang, Victor Zhong, Danqi Chen, Gabor Angeli, and Christopher D. Manning. 2017. https://doi.org/10.18653/V1/D17-1004 Position-aware attention and supervised data improve slot filling . In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Proces...

    2017 doi

  31. [39]

    Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, Zheyan Luo, Zhangchi Feng, and Yongqiang Ma. 2024. http://arxiv.org/abs/2403.13372 Llamafactory: Unified efficient fine-tuning of 100+ language models . In Proceedings of the 62nd Annual Meeting of the Association for Compu...

    2024 arXiv

  32. [40]

    Lingfeng Zhong, Jia Wu, Qian Li, Hao Peng, and Xindong Wu. 2024. https://doi.org/10.1145/3618295 A comprehensive survey on automatic knowledge graph construction . ACM Comput. Surv. , 56(4):94:1--94:62

    2024 doi

  33. [41]

    Wenxuan Zhou and Muhao Chen. 2022. https://aclanthology.org/2022.aacl-short.21 An improved baseline for sentence-level relation extraction . In Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th Internati...

    2022

  34. [42]

    URL: " 'urlintro :=

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

  35. [43]

    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 4, 2026 · model on record in the stance chip above.