Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

GLiREL -- Generalist Model for Zero-Shot Relation Extraction

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

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

Pith's one-line read The paper claims a single-pass encoder model, GLiREL, classifies all relation labels against all entity pairs at once and, after synthetic pretraining, beats GPT-4o and matches prior state of the art on zero-shot relation classification…

desk verdict A practical single-pass zero-shot relation classifier and a reusable synthetic data protocol; the SoTA claim outruns the evidence. read the letter →

arxiv 2501.03172 v1 pith:X5NOD4W7 submitted 2025-01-06 cs.CL cs.AIcs.LG

Jack Boylan , Chris Hokamp , Demian Gholipour Ghalandari This is my paper · ORCID
classification cs.CLcs.AIcs.LG
keywords zero-shotrelationextractionclassificationsyntheticdatagenerationGLiRELFewWiki-ZSLentitypairscoringlabelembeddings
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

The paper introduces GLiREL, a transformer-based model that classifies relations between every pair of named entities in a sentence in a single forward pass, using relation labels supplied as text at inference time. It claims this is an efficient architecture that reaches state-of-the-art zero-shot relation classification, beating GPT-4o on Wiki-ZSL and FewRel for every tested number of unseen relations and matching the previous best model at the hardest setting. The efficiency claim matters because prior methods need one forward pass per entity-pair and label combination, which does not scale to documents with many entities and many candidate labels. A second contribution is a synthetic data protocol: an LLM annotates relation labels on web text, and pretraining on this data before benchmark fine-tuning materially improves accuracy. If these claims hold, zero-shot relation extraction can be done with a single small model rather than a large generative LLM.

What carries the argument

The load-bearing object is the label-entity-pair similarity score $\phi(u,v,t) = \sigma(\kappa_{uv}^\top q_t)$, where $q_t$ is the pooled representation of relation label $t$ and $\kappa_{uv}$ is the representation of entity pair $(u,v)$. All labels and all pairs are encoded in one forward pass through a bidirectional transformer, and the scores fill a matrix that is trained with binary cross-entropy. Optional cross-attention refinement layers let pair and label representations update each other before scoring, and the synthetic pretraining protocol supplies the diverse relation labels that make the zero-shot transfer work.

What would settle it

Re-run TMC-BERT and ZSRE under GLiREL's exact protocol — five random selections of m unseen relations with macro F1 — on Wiki-ZSL and FewRel; if either baseline matches or beats GLiREL at m=15 under that protocol, the state-of-the-art claim fails.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that zero-shot relation classification can be reduced to a similarity-ranking problem between entity-pair representations and relation-label representations in one shared latent space, and that this formulation benefits from the same synthetic-data recipe that advanced zero-shot named entity recognition. The model encodes relation labels and the sentence together with DeBERTa, pools each label to a vector, builds a representation for every ordered pair of entities, and scores all label-pair combinations with a dot product followed by a sigmoid. Training with binary cross-entropy and optional cross-attention refinement layers lets the model classify m unseen relations without any examples of those relations. Reported results include 94.20 F1 on FewRel and 83.28 F1 on Wiki-ZSL at m=5, 87.60 and 83.67 at m=10, and 84.48 and 73.91 at m=15, and the paper states that GLiREL outperforms GPT-4o at every value of m on both datasets. The authors also report that GLiREL degrades less than competitors as m grows, which they attribute to encoding labels and text jointly.

Load-bearing premise

The state-of-the-art claim depends on the baseline numbers in Table 1, which are copied from their original papers, being directly comparable to GLiREL's means over five random selections of unseen relations; if the baselines used a different split or selection protocol, the reported margins would not be meaningful.

Editorial extensions

If this is right

  • If the reported results hold, zero-shot relation classification no longer needs a forward pass per entity pair and candidate label; GLiREL classifies all pairs and labels in a single pass, which the paper measures as roughly 20x more sentences per second than RelationPrompt and TMC-BERT on Wiki-ZSL.
  • The synthetic pretraining recipe — sample web text, have an LLM annotate arbitrary relation labels, discard labels that overlap the benchmark, then fine-tune — becomes a reusable protocol for building zero-shot relation classifiers on new label sets.
  • A practitioner can point one trained model at any relation taxonomy at inference time, since relation labels are supplied as plain text and never fixed during training.
  • Because performance degrades less than baselines as the number of unseen labels m grows, the model is suited to applications with large candidate label sets, subject to the 512-token context limit the authors note.

Reading between the lines

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

  • Editorial inference: the same encoder-plus-synthetic-label recipe could transfer to other open-label structured prediction tasks such as event argument extraction or slot filling, where the bottleneck is also arbitrary label sets.
  • Editorial inference: the paper's state-of-the-art comparison is only as good as the comparability of copied baseline numbers; a head-to-head re-run under identical five-seed random splits would settle whether the reported margins are real.
  • Editorial inference: the authors themselves argue Wiki-ZSL contains distantly supervised labels that cannot be inferred from text, and moving evaluation to text-grounded benchmarks like FewRel 2.0 with a NOTA category would test whether GLiREL's zero-shot accuracy is usable in practice.
  • Editorial inference: since predicted coreference clusters hurt Re-DocRED performance badly, a practical document-level system would pair GLiREL with an external coreference resolver rather than learning the SELF relation end-to-end.
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

3 major / 5 minor

Summary. The paper introduces GLiREL, an encoder-based architecture for zero-shot relation classification that extends the GLiNER span-classification approach to relation extraction: relation labels and entity pairs are encoded in a single Transformer forward pass, and all entity-pair/label combinations are scored by dot-product matching. The authors also contribute a synthetic-data protocol in which Mistral 7B-Instruct annotates relation labels on Fineweb text, with benchmark-overlapping labels discarded. Experiments on Wiki-ZSL and FewRel at m=5, 10, and 15 unseen relations, averaged over five random selections, are reported together with a GPT-4o baseline and a Re-DocRED appendix. The paper claims state-of-the-art zero-shot relation classification performance and attributes large gains to synthetic pretraining.

Significance. If the central claims held, the work would be a useful contribution: a single-pass, non-autoregressive relation classifier that handles multiple entity pairs and arbitrary labels is a practical efficiency improvement over per-pair/label baselines, and the public synthetic dataset and training protocol are reusable assets. The architecture is a straightforward adaptation of GLiNER, but the efficiency comparison in Section 5.1 is concrete and the paper ships code and data. However, the headline 'state-of-the-art' claim is not supported by the paper's own Table 1 in several configurations, and the absence of significance testing and protocol verification makes the remaining SoTA margins fragile. The contribution is better framed as an efficient and competitive zero-shot relation classifier, with synthetic pretraining as a promising training recipe, rather than as a consistently state-of-the-art method.

major comments (3)
  1. [Section 4.3 and Table 1] The claim that GLiREL achieves state-of-the-art results on both Wiki-ZSL and FewRel is contradicted by Table 1 in several configurations. With synthetic pretraining, GLiREL scores 83.28 F1 on Wiki-ZSL at m=5, whereas ZSRE scores 95.46 and TMC-BERT 88.92; at m=10, ZSRE still leads with 86.74 versus GLiREL's 83.67. On FewRel at m=5, ZSRE scores 96.51 versus GLiREL's 94.20. GLiREL is best only at Wiki-ZSL m=15 (73.91 vs. TMC-BERT 73.77) and FewRel m=10 and m=15. The abstract and Section 4.3 should either be revised to state the precise conditions under which GLiREL is state of the art or the claim should be removed.
  2. [Table 1, note on baseline results] The state-of-the-art comparison rests on the unverified assumption that baseline numbers copied from their original papers were produced under the same protocol as GLiREL's: the same random selection of m unseen relations and number of repeats, the same train/test splits, the same macro-averaging, and the same treatment of NO_RELATION. The paper's own protocol in Section 4.2 is five random selections with mean macro F1, but no evidence is given that the cited baselines used this protocol. Given that the decisive margin at Wiki-ZSL m=15 is 0.14 F1, protocol mismatch alone could reverse the conclusion. The authors should either reproduce the baselines under their exact protocol or report the original protocols and clearly qualify the comparison.
  3. [Section 4.3, synthetic pretraining attribution] The statement that pretraining on the synthetically annotated dataset shows 'significant improvement' is not supported by significance testing or controlled ablations. Table 1 reports only point estimates of five-seed means for GLiREL, with no standard deviations, so the improvement from synthetic pretraining (e.g., 62.80 to 83.28 on Wiki-ZSL m=5; 81.21 to 94.20 on FewRel m=5) cannot be separated from optimization variance or seed effects. Additionally, GLiREL uses DeBERTa-large (467M parameters) while TMC-BERT has 109M, so part of the gap may be attributable to model capacity. The authors should report variance across seeds and, ideally, ablate pretraining data scale or use a smaller encoder to isolate the contribution of the synthetic pretraining protocol.
minor comments (5)
  1. [Abstract] The abstract says 'achieves state-of-the-art results on the zero-shot relation classification task' without qualification; this should be aligned with the configuration-specific results in Table 1.
  2. [Figures] There are two figures numbered 'Figure 1' (the example inputs and the architecture diagram), which makes cross-references in Section 3 confusing and should be corrected.
  3. [Various] There are several typos and formatting issues: 'langage' in Section 3, 'Due of the difficulty' in Section 3.6, 'at every value ofm' in Section 4.3, 'approache' in Appendix A.7, and 'archictecture' in the same appendix.
  4. [Appendix A.3] The heading for Figure 7 says 'Prompt for synthetic dataset generation', but the figure shows the GPT-4o baseline prompt; this is likely a copy-paste error and should be corrected.
  5. [Table 1] Reporting standard deviations for the five random selections would substantially improve the interpretability of the results, especially for the small margins at Wiki-ZSL m=15.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: this is an empirical benchmark study with disjoint held-out labels; the main risks are baseline comparability and variance, not self-referential reasoning.

full rationale

The paper contains no formal derivation chain whose output could reduce to its inputs. All central claims are empirical comparisons on the FewRel and Wiki-ZSL benchmarks. The model is trained with binary cross-entropy and scores relation candidates by a dot product between learned entity-pair and label representations (Eqs. 7-8); this is an architecture choice, not a definitional equivalence between a predicted quantity and a fitted input. The synthetic pretraining intervention is externally sourced: Mistral 7B annotates Fineweb texts, and Section 3.6 explicitly states that labels intersecting with benchmark labels are discarded 'in order to strictly maintain the zero-shot paradigm.' Therefore the benchmark evaluation labels are not reused in pretraining, and the reported gains are not forced by construction. The strongest substantive concern is that Table 1 compares GLiREL's five-seed means with baseline numbers 'copied from their original papers,' so protocol equivalence is unverified; but that is a validity and comparability threat, not circularity. Similarly, the absence of significance tests and standard deviations is a statistical reporting weakness, not a self-referential step. The architectural debt to GLiNER and NuNER is openly acknowledged and refers to external prior work rather than the authors' own unverified results, so no self-citation chain is load-bearing. The limitations section honestly identifies real issues such as sequence-length constraints and Wiki-ZSL distant-supervision noise, but none of these indicate that a prediction was equivalent to its input. Overall, no concrete circular step can be quoted from the paper, and the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The paper's contributions are empirical. The central claims rest on standard ML background assumptions: the pretrained DeBERTa encoder, the quality of LLM-generated synthetic relation annotations, the validity of the distant-supervision benchmarks, and the completeness of label-overlap removal between synthetic and benchmark data. The paper's own Limitations section notes that Wiki-ZSL can contain relations unanswerable from text alone. No new scientific entities are introduced.

free parameters (2)
  • Max relation labels per training instance = 25
    Section 4.2: 'We limit the number of relation type labels prepended to each training instance to 25.' Chosen by hand; affects training distribution and model capacity.
  • Number of refinement layers = 2 (maximum)
    Section 3.4: 'in practice we use a maximum of two refinement layers for efficiency.' Architecture choice; the ablation shows mixed effects across datasets.
assumptions (4)
  • domain assumption DeBERTa V3-large pretrained representations are a suitable encoder for joint label and text encoding.
    Section 3: 'We choose DeBERTa V3-large as the encoder model due to its excellent performance on downstream tasks.' Relies on prior pretraining; model weights are external.
  • domain assumption Mistral 7B-Instruct annotates relation labels accurately and diversely in the synthetic dataset.
    Section 3.6: 'We use Mistral 7B-Instruct-v0.3 to annotate every entity pair in every text.' The quality of the synthetic data is assumed; the paper does not measure annotation agreement.
  • domain assumption Wikipedia-derived distant supervision (Wiki-ZSL, FewRel) provides valid ground-truth relations for evaluation.
    Section 4.1 describes both datasets as generated through distant supervision; Section 7 itself notes Wiki-ZSL texts often lack evidence for the imputed relation.
  • domain assumption Benchmark relation labels in the synthetic data are fully identified and discarded.
    Section 3.6: 'we discard those labels that intersect with benchmark labels.' Requires complete overlap detection between LLM-generated labels and the benchmark label sets.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GLiREL -- Generalist Model for Zero-Shot Relation Extraction." pith.science (2026). https://pith.science/paper/X5NOD4W7

@misc{pith2026250103172,
  author       = {Pith},
  title        = {Pith review of: GLiREL -- Generalist Model for Zero-Shot Relation Extraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X5NOD4W7}},
  note         = {Machine review of arXiv:2501.03172}
}
read the original abstract

We introduce GLiREL (Generalist Lightweight model for zero-shot Relation Extraction), an efficient architecture and training paradigm for zero-shot relation classification. Inspired by recent advancements in zero-shot named entity recognition, this work presents an approach to efficiently and accurately predict zero-shot relationship labels between multiple entities in a single forward pass. Experiments using the FewRel and WikiZSL benchmarks demonstrate that our approach achieves state-of-the-art results on the zero-shot relation classification task. In addition, we contribute a protocol for synthetically-generating datasets with diverse relation labels.

Figures

Figures reproduced from arXiv: 2501.03172 by the authors.

Figure 1
Figure 1. Example inputs and outputs for GLiREL. formance, but are inefficient because every entity pair and candidate label combination is treated as a separate input. Existing methods do not scale to real-world use cases, where a large number of entity pairs is extracted from text, each of which must be classified against many candidate labels. GLiREL takes inspiration from recent successes in zero-shot NER and text classif… view at source ↗
Figure 2
Figure 2. Our proposed approach to zero-shot relation extraction. Firstly, the relation labels and N entities are [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 5
Figure 5. Addition of refinement layers: The effect of adding refine layers for entity pair and relation labels representations. From the WikiZSL dataset, using m = 15 70 72 74 76 78 80 82 84 86 Mean Macro F1 Refine prompt + relation Refine prompt only Refine relation only No refinement 84.48 80.98 77.43 81.29 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figures from the paper (5 more)
Figure 6
Figure 6. Figure 6: Addition of refinement layers: The effect of adding refine layers for entity pair and relation labels representations. From the FewRel dataset, using m = 15 6 Conclusion We have shown that GliREL is a flexible and highly performant approach to zero-shot relation clas￾s…
Figure 4
Figure 4. Figure 4: Addition of random drop: The effect of randomly dropping relation labels during training on the FewRel and WikiZSL datasets. Using m = 15. Refinement Layers The refinement layers as de￾scribed in Section 3.4 aim to enhance the represen￾tations of both the entity pair r…
Figure 8
Figure 8. Figure 8: Prompt for synthetic dataset generation. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 7
Figure 7. Figure 7: Prompt for synthetic dataset generation. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 9
Figure 9. Figure 9: An example of merging entities into clusters [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A Tale of LLMs and Induced Small Proxies: Scalable Small Language Models for Knowledge Mining

    cs.AI 2025-10 conditional novelty 6.0 of 10

    LLM-written pipelines and LLM-generated labels are distilled into one small instruction-following model that performs classification and span extraction cheaply at corpus scale.

Reference graph

Works this paper leans on

43 extracted references · 24 canonical work pages · cited by 1 Pith paper

  1. [6]

    Lifted Rule Injection for Relation Embeddings

    Lifted rule injection for relation em- beddings. Preprint, arXiv:1606.08359. Markus Eberts and Adrian Ulges

    work page Pith review arXiv

  2. [8]

    Preprint, arXiv:2112.04539

    Prompt-based zero-shot relation extraction with semantic knowl- edge augmentation. Preprint, arXiv:2112.04539. Xu Han, Hao Zhu, Pengfei Yu, Ziyun Wang, Yuan Yao, Zhiyuan Liu, and Maosong Sun

    arXiv

  3. [11]

    Preprint, arXiv:2310.06825

    Mistral 7b. Preprint, arXiv:2310.06825. Yizhi Jiang, Jinlong Li, and Huanhuan Chen

    arXiv

  4. [12]

    In Proceedings of the 2024 Joint International Con- ference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING

    Re- lation classification via bidirectional prompt learn- ing with data augmentation by large language model. In Proceedings of the 2024 Joint International Con- ference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING

    work page 2024

  5. [13]

    In 2023 In- ternational Joint Conference on Neural Networks (IJCNN), pages 1–8

    Modeling zero-shot relation clas- sification as a multiple-choice problem. In 2023 In- ternational Joint Conference on Neural Networks (IJCNN), pages 1–8. Omer Levy, Minjoon Seo, Eunsol Choi, and Luke Zettle- moyer

    work page 2023

  6. [15]

    Preprint, arXiv:2310.05028

    Re- visiting large language models as zero-shot relation extractors. Preprint, arXiv:2310.05028. Guozheng Li, Peng Wang, Jiajun Liu, Yikai Guo, Ke Ji, Ziyu Shang, and Zijie Xu. 2024a. Meta in- context learning makes large language models bet- ter zero and few-shot relation extractors. Preprint, arXiv:2404.17807. Xingzuo Li, Kehai Chen, Yunfei Long, and Mi...

    arXiv

  7. [16]

    In Findings of the Association for Computational Linguistics: ACL 2023, pages 5491–5505, Toronto, Canada

    DSP: Discriminative soft prompts for zero-shot entity and relation extraction. In Findings of the Association for Computational Linguistics: ACL 2023, pages 5491–5505, Toronto, Canada. Association for Computational Linguistics. Youmi Ma, An Wang, and Naoaki Okazaki. 2023a. DREEAM: Guiding attention with evidence for im- proving document-level relation ext...

    work page 2023

  8. [17]

    Preprint, arXiv:2406.17557

    The fineweb datasets: Decanting the web for the finest text data at scale. Preprint, arXiv:2406.17557. Colin Raffel, Noam Shazeer, Adam Roberts, Kather- ine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu

    arXiv

Show all 43 references
  1. [19]

    Preprint, arXiv:2104.08481

    Revisiting few-shot relation classification: Evaluation data and classification schemes. Preprint, arXiv:2104.08481. Oscar Sainz, Iker García-Ferrero, Rodrigo Agerri, Oier López de Lacalle, German Rigau, and Eneko Agirre

    arXiv

  2. [20]

    ArXiv, abs/2310.03668

    Gollie: Annotation guidelines im- prove zero-shot information-extraction. ArXiv, abs/2310.03668. Oscar Sainz, Oier Lopez de Lacalle, Gorka Labaka, Ander Barrena, and Eneko Agirre

    arXiv

  3. [21]

    In Proceedings of the 2021 Conference on Empirical Methods in Natural Lan- guage Processing, pages 1199–1212, Online and Punta Cana, Dominican Republic

    Label ver- balization and entailment for effective zero and few- shot relation extraction. In Proceedings of the 2021 Conference on Empirical Methods in Natural Lan- guage Processing, pages 1199–1212, Online and Punta Cana, Dominican Republic. Association for Computational Lin...

    2021

  4. [22]

    In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 1784–1789, Copenhagen, Denmark

    Context- aware representations for knowledge base relation extraction. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 1784–1789, Copenhagen, Denmark. Associa- tion for Computational Linguistics. Ihor Stepanov and Mykhailo Shtopko

    2017

  5. [23]

    Preprint, arXiv:2406.12925

    GLiNER multi-task: Generalist lightweight model for var- ious information extraction tasks. Preprint, arXiv:2406.12925. Qi Sun, Kun Huang, Xiaocui Yang, Rong Tong, Kun Zhang, and Soujanya Poria

    arXiv

  6. [24]

    Preprint, arXiv:2401.13598

    Consistency guided knowledge retrieval and denoising in llms for zero-shot document-level relation triplet extraction. Preprint, arXiv:2401.13598. Qingyu Tan, Ruidan He, Lidong Bing, and Hwee Tou Ng

    arXiv

  7. [25]

    In Findings of the Association for Computational Lin- guistics: ACL 2022, pages 1672–1681, Dublin, Ire- land

    Document-level relation extraction with adaptive focal loss and knowledge distillation. In Findings of the Association for Computational Lin- guistics: ACL 2022, pages 1672–1681, Dublin, Ire- land. Association for Computational Linguistics. Qingyu Tan, Lu Xu, Lidong Bing, Hwee...

    2022

  8. [26]

    Preprint, arXiv:2205.12696

    Revisiting Do- cRED – addressing the false negative problem in relation extraction. Preprint, arXiv:2205.12696. Van-Hien Tran, Hiroki Ouchi, Hiroyuki Shindo, Yuji Matsumoto, and Taro Watanabe

    arXiv

  9. [27]

    Preprint, arXiv:1706.03762

    Attention is all you need. Preprint, arXiv:1706.03762. Denny Vrandeˇci´c and Markus Krötzsch

    arXiv

  10. [28]

    In Findings of the Association for Computational Lin- guistics: ACL 2022, pages 803–823, Dublin, Ireland

    DeepStruct: Pre- training of language models for structure prediction. In Findings of the Association for Computational Lin- guistics: ACL 2022, pages 803–823, Dublin, Ireland. Association for Computational Linguistics. Xin Xu, Yuqi Zhu, Xiaohan Wang, and Ningyu Zhang

    2022

  11. [29]

    Lilong Xue, Dan Zhang, Yuxiao Dong, and Jie Tang

    How to unleash the power of large language models for few-shot relation extraction? Preprint, arXiv:2305.01555. Lilong Xue, Dan Zhang, Yuxiao Dong, and Jie Tang

    arXiv

  12. [30]

    Preprint, arXiv:2403.14888

    Autore: Document-level relation ex- traction with large language models. Preprint, arXiv:2403.14888. Deming Ye, Yankai Lin, Jiaju Du, Zhenghao Liu, Peng Li, Maosong Sun, and Zhiyuan Liu

    arXiv

  13. [32]

    Preprint, arXiv:2404.12493

    EnriCo: En- riched representation and globally constrained in- ference for entity and relation extraction. Preprint, arXiv:2404.12493. Urchade Zaratiana, Nadi Tomeh, Pierre Holat, and Thierry Charnois

    arXiv

  14. [33]

    Preprint, arXiv:2311.08526

    GLiNER: Generalist model for named entity recognition using bidirectional trans- former. Preprint, arXiv:2311.08526. 11 Yuhao Zhang, Victor Zhong, Danqi Chen, Gabor Angeli, and Christopher D. Manning

    arXiv

  15. [34]

    In Proceedings of the 2017 Conference on Empiri- cal Methods in Natural Language Processing, pages 35–45, Copenhagen, Denmark

    Position-aware attention and supervised data improve slot filling. In Proceedings of the 2017 Conference on Empiri- cal Methods in Natural Language Processing, pages 35–45, Copenhagen, Denmark. Association for Com- putational Linguistics. Jun Zhao, WenYu Zhan, Xin Zhao, Qi Zha...

    2017

  16. [35]

    Preprint, arXiv:2308.03279

    Universalner: Targeted dis- tillation from large language models for open named entity recognition. Preprint, arXiv:2308.03279. A Appendix A.1 Extended Related Work Existing Approaches Many systems have ad- dressed relation extraction with varying degrees of success. Earlier w...

    2022 arXiv

  17. [37]

    participation in

    dataset. The authors’ experiments reveal that cur- rent state-of-the-art models and techniques strug- gle with these additional challenges that more ac- curately mirror real-world application of relation extraction models. A.2 Tokenization Details The special [REL] and [SEP] t...

    2022

  18. [39]

    Malala Yousafzai, the youngest Nobel Peace Prize laureate, gave an inspiring speech at the United Nations. Later, Yousafzai met with Justin Trudeau to discuss education reform

    Initial Learning Rate (Encoder) 1 × 10−5 Initial Learning Rate (Other Parameters) 1 × 10−4 Warmup Ratio 10% Scheduler Cosine Hidden Layer Size (Non-encoder Layers) 768 Batch Size 8 Total Training Steps 20,000 GPU NVIDIA Tesla T4 "Malala Yousafzai, the youngest Nobel Peace Priz...

    2023

  19. [40]

    Ma et al

    achieves SoTA results using a RoBERTa-based model, with the addition of an axial attention module to capture interdependen- cies among entity pairs, and a knowledge distil- lation framework to make use of large-scale dis- tantly supervised data. Ma et al. (2023a) perform stron...

    2023

  20. [41]

    (2022) 87.48 77.50 82.19 87.11 86.29 86.69 RelationPrompt NG (Chia et al.,

    71.54 72.39 71.96 76.96 78.86 77.90 MICRE w/Llama (Li et al., 2024a) 76.46 78.53 77.48 89.34 91.88 90.59 Tran et al. (2022) 87.48 77.50 82.19 87.11 86.29 86.69 RelationPrompt NG (Chia et al.,

    2022

  21. [42]

    (2022) 71.59 64.69 67.94 64.41 62.61 63.50 RelationPrompt NG (Chia et al.,

    60.51 60.98 60.74 56.92 57.59 57.25 MICRE w/Llama (Li et al., 2024a) 72.36 74.88 73.60 80.67 82.31 81.48 Tran et al. (2022) 71.59 64.69 67.94 64.41 62.61 63.50 RelationPrompt NG (Chia et al.,

    2022

  22. [43]

    (2022) 38.37 36.05 37.17 43.96 39.11 41.36 RelationPrompt NG (Chia et al.,

    34.12 34.38 34.25 35.54 38.19 36.82 MICRE w/Llama (Li et al., 2024a) 67.14 68.87 67.99 73.74 75.83 74.77 Tran et al. (2022) 38.37 36.05 37.17 43.96 39.11 41.36 RelationPrompt NG (Chia et al.,

    2022

  23. [768]

    All experiments were carried out using one NVIDIA Tesla T4 GPU

    The batch size was 8, with total number of steps set to 20,000. All experiments were carried out using one NVIDIA Tesla T4 GPU. A.6 Full Zero-Shot Relation Classification Results A.7 Coreference Resolution and Document-level Relation Classification We conceptualize coreference...

    2023

  24. [2007]

    none-of-the-above

    to obtain high-quality relations be- tween entities, and then mining data sources such as Wikipedia for texts mentioning both head and tail entities to construct training datasets (Bunescu and Mooney, 2007; Mintz et al., 2009). Distant supervision enables the creation of large...

    2021

  25. [2008]

    In Proceedings of the 2008 ACM SIGMOD International Conference on Management of Data, SIGMOD ’08, page 1247–1250, New York, NY , USA

    Freebase: a col- laboratively created graph database for structuring human knowledge. In Proceedings of the 2008 ACM SIGMOD International Conference on Management of Data, SIGMOD ’08, page 1247–1250, New York, NY , USA. Association for Computing Machinery. Razvan Bunescu and R...

    2008

  26. [2013]

    In Pro- ceedings of the 2013 Conference of the North Amer- ican Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 74–84, Atlanta, Georgia

    Relation extraction with matrix factorization and universal schemas. In Pro- ceedings of the 2013 Conference of the North Amer- ican Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 74–84, Atlanta, Georgia. Association for Computa- t...

    2013 arXiv

  27. [2016]

    Comparing convolutional neural networks to tradi- tional models for slot filling. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Hu- man Language Technologies, pages 828–838, San Diego, California. Associat...

    2016

  28. [2017]

    Preprint, arXiv:1706.04115

    Zero-shot relation extraction via read- ing comprehension. Preprint, arXiv:1706.04115. Guozheng Li, Peng Wang, and Wenjun Ke

    arXiv

  29. [2018]

    Preprint, arXiv:1810.10147

    Fewrel: A large-scale supervised few-shot relation classifica- tion dataset with state-of-the-art evaluation. Preprint, arXiv:1810.10147. Pengcheng He, Jianfeng Gao, and Weizhu Chen

    arXiv

  30. [2019]

    FewRel 2.0: Towards more challenging few-shot relation classi- fication. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natu- ral Language Processing (EMNLP-IJCNLP), pages 6250–6255, Hong ...

    2019

  31. [2020]

    In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7170–7186, Online

    Corefer- ential Reasoning Learning for Language Represen- tation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7170–7186, Online. Association for Computational Linguistics. Urchade Zaratiana, Nadi Tomeh, Yann Dauxais,...

    2020

  32. [2021]

    Preprint, arXiv:2104.04697

    ZS-BERT: Towards zero-shot relation extraction with attribute representation learning. Preprint, arXiv:2104.04697. Yew Ken Chia, Lidong Bing, Soujanya Poria, and Luo Si

    arXiv

  33. [2022]

    In Findings of the Association for Compu- tational Linguistics: ACL 2022, pages 45–57, Dublin, Ireland

    RelationPrompt: Leveraging prompts to generate synthetic data for zero-shot relation triplet extraction. In Findings of the Association for Compu- tational Linguistics: ACL 2022, pages 45–57, Dublin, Ireland. Association for Computational Linguistics. Thomas Demeester, Tim Roc...

    2022

  34. [2023]

    Preprint, arXiv:2111.09543

    Debertav3: Improving deberta using electra-style pre- training with gradient-disentangled embedding shar- ing. Preprint, arXiv:2111.09543. Matthew Honnibal, Ines Montani, Sofie Van Lan- deghem, and Adriane Boyd

    arXiv

  35. [2024]

    Preprint, arXiv:2402.15343

    NuNER: Entity recognition encoder pre-training via LLM-annotated data. Preprint, arXiv:2402.15343. Kurt Bollacker, Colin Evans, Praveen Paritosh, Tim Sturge, and Jamie Taylor

    arXiv

Pith tools

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