Pith. sign in

REVIEW 4 major objections 6 minor 63 references

CPTuning: Contrastive Prompt Tuning for Generative Relation Extraction

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read CPTuning claims that contrastive prompt tuning with layer-based label smoothing lets a sequence-to-sequence model assign each candidate relation a probability above or below a threshold, so thresholding generated candidates extracts all…

desk verdict Solid multi-relation RE results, but the core threshold objective and inference rule are written backwards. read the letter →

arxiv 2501.02196 v1 pith:JJ6XN2GM submitted 2025-01-04 cs.CL cs.AI

Jiaxin Duan , Fengyu Lu , Junfei Liu This is my paper · ORCID
classification cs.CLcs.AI
keywords relationextractionentitypairoverlapcontrastivelearningprompttuningsequence-to-sequencegenerationtextinfillinglabelsmoothingTrie-constraineddecoding
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

Relation extraction traditionally assigns exactly one relation to each entity pair, but real sentences often state several relations at once; the paper calls this entity pair overlap. CPTuning is a training and decoding scheme that lets a sequence-to-sequence model output all valid relations instead of one. The central claim is that contrastive training with layer-based label smoothing makes the model associate each candidate relation with probability mass above or below a threshold, and that thresholding the generated candidates recovers every true relation. The authors report that T5-large fine-tuned with CPTuning outperforms previous generative and classificational methods on four datasets, covering both single- and multi-relation extraction.

What carries the argument

The central object is a thresholded probability mass over verbalized relation phrases. The training objective is $L(\theta) = L_{\mathrm{ctl}}(\theta) + \mu L_{\mathrm{lbls}}(\theta)$, where $L_{\mathrm{lbls}}$ is the layer-based label smoothing loss that distributes target probability $1-\beta$ to the gold token and $\beta/(|T_i|-1)$ to the other tokens allowed at layer $i$ of the Trie, and $L_{\mathrm{ctl}}$ is a contrastive loss that penalizes gold-relation scores $f(r_i)$ below $\zeta$ and fake-relation scores above $\zeta$. The score $f$ is a length-normalized negative log-likelihood. Decoding uses a Trie, a prefix tree over the verbalized relation vocabulary, to restrict beam search to valid relation phrases, and the final threshold $\lambda$ converts the ranked candidate list into the extracted relation set.

What would settle it

Manually inspect a sample of the fake negative relations used in Eq. 7 on the NYT training set: if a meaningful fraction of them are actually expressed in the sentence, then the contrastive loss is suppressing true relations, and replacing random negatives with verified or entity-type-incompatible negatives should raise multi-relation F1 above the reported 91.2-91.4.

Watch

Extended reading notes

Core claim

The paper claims that entity pair overlap can be handled by a generative model that learns a graded, threshold-able likelihood over relation phrases rather than a one-hot gold label. Concretely, CPTuning combines layer-based label smoothing, which gives every valid token at each decoding layer a small nonzero target probability, with a contrastive loss that pushes gold relations above a threshold $\zeta$ and randomly sampled fake relations below it. At inference, Trie-constrained prefix-given beam search generates a set of candidate relations and a length-normalized log-likelihood score $f(r_i)$ is compared with a threshold $\lambda$ to decide which relations exist. The paper reports that T5-large trained this way reaches the best micro F1 among compared methods on TACRED, TACREV and NYT, with Re-TACRED used in ablations and low-resource experiments, outperforming both classificational and generative baselines in single- and multi-relation settings.

Load-bearing premise

The load-bearing premise is that a relation sampled at random from the relation vocabulary and not labelled for a given entity pair is genuinely absent; on automatically annotated data such as NYT this can fail because a sampled relation may be true but simply unannotated.

Editorial extensions

If this is right

  • On NYT, a dataset built for entity pair overlap, CPTuning with T5-large reaches micro F1 of 91.4 while the strongest prior generative prompt-tuning baseline reaches 87.27, so thresholded generative decoding works for multi-relation extraction.
  • On single-relation datasets, the same training recipe beats the previous best generative method by 1.8 F1 points on TACRED and 0.9 on TACREV, so supporting multiple relations does not degrade single-relation accuracy.
  • Ablation removes either the contrastive loss or the layer-based label smoothing and NYT F1 drops from 91.2 to about 87.7-87.9, indicating both components are needed for the reported multi-relation gain.
  • With only 8-32 labeled instances per relation, CPTuning is competitive with or better than prior prompt-tuning baselines on most settings, suggesting the thresholded likelihood signal remains useful in low-resource conditions.

Reading between the lines

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

  • The paper leaves implicit that the same contrastive thresholding recipe could apply to other structured generation problems with variable numbers of valid outputs per input, such as event extraction or open information extraction, since the machinery only requires a verbalized candidate set and a decision threshold.
  • Because the negative relations in Eq. 7 are sampled uniformly from the relation vocabulary, the method's practical ceiling on distantly supervised data may be set by annotation noise; an obvious test is to replace uniform negatives with entity-type-constrained or confidence-weighted negatives and check whether multi-relation F1 rises.
  • The H-index analysis in the paper suggests likelihood and semantic similarity align under CPTuning; one concrete extension is to use semantic similarity rankings as an additional training signal, which could reduce the number of beam candidates needed at inference.
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

4 major / 6 minor

Summary. The paper proposes CPTuning, a contrastive prompt-tuning method for generative relation extraction. It reformulates RE as a Seq2Seq text-infilling task with sentinel tokens, uses Trie-constrained beam search to generate candidate verbalized relations, and combines layer-based label smoothing (LBLS) with a contrastive loss (CTL) to let the model assign probability mass above or below a threshold. Experiments on TACRED, TACREV, Re-TACRED, and NYT report state-of-the-art micro-F1 scores for T5-large, with ablations showing that both LBLS and CTL contribute to the gains.

Significance. If the method works as described, it addresses a real limitation of generative prompt-tuning RE—entity-pair overlap—and the reported gains over GenPT and other baselines are substantial. The paper's ablations (Table 3) support the contribution of both proposed components, and the H-index analysis (Table 4) is a useful diagnostic of semantic behavior. However, the central scoring and threshold mechanism is written inconsistently, so the claimed results cannot be verified from the text; no code is provided.

major comments (4)
  1. [§3.2, §3.3, Eq. (3), Eq. (7), Algorithm 1] Eq. (3) defines f(ri) = −Σ log P(ri(j)) / |ri|^α, so larger f means lower token probability. Eq. (7) then minimizes max(ζ − f(ri), 0) for gold relations, which is zero only when f(ri) ≥ ζ (i.e., low probability), and max(f(rj) − ζ, 0) for fake relations, which is zero only when f(rj) ≤ ζ (i.e., high probability). Algorithm 1 line 14 keeps candidates with f(ri) > λ, again selecting low-probability candidates. Taken literally, the training objective and inference rule suppress gold relations and promote fake ones, so the reported F1 gains in Table 2 cannot follow from the written algorithm. Please correct the sign of f or flip the inequalities, and clarify which convention was actually implemented.
  2. [§4.4, Table 2] Results in Table 2 are reported as single micro-F1 numbers without error bars, significance tests, or the number of seeds, despite the abstract's claim of "significantly outperforms." The authors should either report multiple runs with variance and a statistical test or soften the significance claim.
  3. [§5, Figure 5] Figure 5 shows performance of CPTuning-s1 and -s2 against μ on TACRED and NYT, and μ = 0.1 selected from these curves is then fixed for all experiments, including TACREV and Re-TACRED. If these curves are computed on the test splits, this is tuning on the test set; please specify the split used and report sensitivity to λ and ζ, which are otherwise fixed without analysis.
  4. [§3.3, Eq. (7)] Equation (7) samples fake relations R− uniformly from the verbalized relation vocabulary and treats them as definitely absent. In a distantly supervised multi-relation dataset such as NYT, an unannotated relation can be true for a given entity pair, so random negative sampling can suppress correct relations and undermine the EPO handling the method targets. Please discuss this issue and consider filtering negatives by entity-type constraints or evaluating alternative negative-sampling strategies.
minor comments (6)
  1. [§5, Figure 5] The manuscript contains template artifacts: Section 5 includes "2 F. Author et al." and Springer template captions (e.g., "Fig. 1: Caption") immediately before Figure 5; these should be removed.
  2. [§4.4 and Figure 4] There are typos: "ralation" in the Section 4.4 headings, "citiy" in Figure 4, and "T able" in table captions.
  3. [§3.2, Algorithm 1] Algorithm 1 initializes r1..rK as the prefix z, but candidate relations should arguably be continuations after the prefix; please clarify how full target texts are represented.
  4. [General] No code or reproducibility statement is provided; given the sign ambiguity in Eq. (3)/(7), releasing code would materially help verification.
  5. [§3.3, Eq. (6)] Equation (6) uses r*(i) both as the predicted token and as the summation index; please clean up the notation to avoid confusion.
  6. [§4.1, Table 1] The text says the NYT training and test sets contain 9782 and 987 EPO entity pairs, but Table 1 only lists "Yes" in the EPO column; please make the table consistent with the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: reported gains are external-benchmark measurements with standard hyperparameter tuning; the only notable issue is an internal sign inconsistency, not circularity.

full rationale

The paper's central claims are empirical comparisons on four external benchmarks (TACRED, TACREV, Re-TACRED, NYT), and the reported F1 numbers are new measurements of a proposed training objective rather than quantities derived from the method's own assumptions. The contrastive loss (Eq. 7) and scoring rule (Eq. 3) are not equivalent to the training labels by construction: they define a learned thresholding scheme whose effectiveness is tested in Table 3, including ablations showing that adding LBLS and CTL changes performance. Hyperparameters such as alpha, beta, lambda, zeta, and mu are manually set or selected via the Figure 5 sweep; choosing a balance factor on development data is standard model selection, not a fitted parameter renamed as a prediction. The method builds on prior external work (GenPT [11], SURE [28], BRIO [26]) rather than on a self-citation chain, and no uniqueness claim is imported from the authors' own prior results. A separate correctness concern, outside circularity, is that Eq. 3 defines f as length-normalized negative log-likelihood, so smaller f corresponds to higher token probability, while Eq. 7 and Algorithm 1 treat larger f as more favorable; this is an internal sign inconsistency that makes the written algorithm hard to reproduce as stated, but it does not make the derivation circular because the reported results are not forced by the equations alone.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central method is an integration of known learning components (label smoothing, contrastive loss, Trie-constrained beam search) and a threshold-based inference rule. The free parameters are hyperparameters hand-set or tuned on the evaluation datasets. The most fragile assumptions are the cleanliness of randomly sampled negatives and the global calibratability of the threshold, both of which are load-bearing for the claimed multi-relation performance.

free parameters (6)
  • alpha (length penalty in Eq. 3) = 0.6
    Hand-selected length normalization exponent for relation scoring; no sensitivity analysis is shown.
  • beta (label smoothing weight in Eq. 6) = 0.2
    Hand-selected smoothing mass for LBLS; no sensitivity analysis is shown.
  • lambda (relation existence threshold, Algorithm 1) = 1.0
    Hand-set cutoff on f; no calibration analysis across datasets, and the direction of the comparison in Algorithm 1 conflicts with Eq. 7.
  • zeta (contrastive margin in Eq. 7) = 1.2
    Hand-set margin separating gold and fake relation scores.
  • mu (loss balance factor in Eq. 8) = 0.1
    Tuned on TACRED and NYT via Figure 5; the text does not state whether the curves are dev or test.
  • K (beam size and number of sampled relations) = 16
    Beam width for prefix-given constrained decoding and number of gold/fake relations sampled in the contrastive loss.
assumptions (4)
  • domain assumption Verbalized relation phrases preserve enough semantics for a T5 model to generate and score relations.
    The whole method relies on the verbalizer v(.) mapping labels to short phrases; if the phrases are ambiguous, generation and threshold scoring degrade. Section 3.1 introduces the verbalizer.
  • domain assumption Randomly sampled relations from v(Y) are genuinely absent for the entity pair during contrastive training.
    Eq. 7 treats R^-_s as fake without checking annotation completeness; distantly supervised datasets like NYT have incomplete labels, so random negatives may be true relations.
  • ad hoc to paper A single global threshold lambda (and margin zeta) can separate true from false relation scores across all entity pairs after label smoothing.
    No calibration analysis is provided; f is a length-normalized negative log-likelihood whose scale varies with model confidence and dataset.
  • standard math All candidate relations for an instance share an identical prefix z, so P(z|M(s)) can be dropped from Eq. 1.
    Holds by construction of the two templates in Figure 2, but it is still a structural assumption that the template prefix carries no usable scoring information. Section 3.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CPTuning: Contrastive Prompt Tuning for Generative Relation Extraction." pith.science (2026). https://pith.science/paper/JJ6XN2GM

@misc{pith2026250102196,
  author       = {Pith},
  title        = {Pith review of: CPTuning: Contrastive Prompt Tuning for Generative Relation Extraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JJ6XN2GM}},
  note         = {Machine review of arXiv:2501.02196}
}
read the original abstract

Generative relation extraction (RE) commonly involves first reformulating RE as a linguistic modeling problem easily tackled with pre-trained language models (PLM) and then fine-tuning a PLM with supervised cross-entropy loss. Although having achieved promising performance, existing approaches assume only one deterministic relation between each pair of entities without considering real scenarios where multiple relations may be valid, i.e., entity pair overlap, causing their limited applications. To address this problem, we introduce a novel contrastive prompt tuning method for RE, CPTuning, which learns to associate a candidate relation between two in-context entities with a probability mass above or below a threshold, corresponding to whether the relation exists. Beyond learning schema, CPTuning also organizes RE as a verbalized relation generation task and uses Trie-constrained decoding to ensure a model generates valid relations. It adaptively picks out the generated candidate relations with a high estimated likelihood in inference, thereby achieving multi-relation extraction. We conduct extensive experiments on four widely used datasets to validate our method. Results show that T5-large fine-tuned with CPTuning significantly outperforms previous methods, regardless of single or multiple relations extraction.

Figures

Figures reproduced from arXiv: 2501.02196 by the authors.

Figure 1
Figure 1. The standard paradigm of relations extraction. example, classificational FPC [53] reformats RE as relation words prediction to align with the masked language modeling (MLM) pre-training objective [9], while generative GenPT [11] converts RE to text-infilling to coordinate with sequence-to-sequence (Seq2Seq) MLM [18]. Compared with traditional methods that predict numerical relation labels depending on contextual fea… view at source ↗
Figure 2
Figure 2. The syntax structure of (a) our handcrafted templates and (b) the synthetic source and target texts in the transformed RE task. {sentence}: instance text slot, {entX}: entity slots, {relation}: relation words slot, and {typeX}: entity type slots. Sentinel tokens [X], [Y], and [Z] corrupt an original formalized instance to build a source text. [E] indicates the end of a target text. that align downstream tasks with t… view at source ↗
Figure 3
Figure 3. The illustration of the Seq2Seq text-infilling task transformed from RE. highlighted entities e i 1 and e i 2 , the objective of RE is to predict these two entities’ relation yi ∈ Y. Existing works always view RE as a multi-class classification (without considering EPO) or multi-label classification (consider EPO) task, modeling the label distribution P(y|xi , ei 1 , ei 2 ), y ∈ Y. Intuitively, any classifier like m… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The illustration of Trie, PGC decoding (a), and LBLS (b). The five relations predefined beforehand are ”per:city of birth”, ”per:citiy of residence”, ”per:city of death”, ”per:country of birth”, and ”per:country of death”, respectively. We contrast the probability mass…
Figure 1
Figure 1. Figure 1: Caption Displayed equations are centered and set on a separate line. x + y = z (1) Please try to avoid rasterized images for line-art diagrams and schemas. When￾ever possible, use vector graphics instead (see [PITH_FULL_IMAGE:figures/full_fig_p013_1.png]
Figure 12
Figure 12. Figure 12 [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 2
Figure 2. Figure 2: A figure caption is always placed below the illustration. Please note that Effect of the Balance Factor. To test the influence of the balance factor [PITH_FULL_IMAGE:figures/full_fig_p013_2.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 55 canonical work pages

  1. [1]

    In: ACL 2020

    Alt, C., Gabryszak, A., Hennig, L.: TACRED revisited: A thorough evaluation of the TACRED relation extraction task. In: ACL 2020. pp. 1558–1569 (2020)

    work page 2020

  2. [2]

    In: EMNLP-IJCNLP 2019

    Balazevic, I., Allen, C., Hospedales, T.M.: Tucker: Tensor factorization for knowl- edge graph completion. In: EMNLP-IJCNLP 2019. pp. 5184–5193 (2019)

    work page 2019

  3. [3]

    In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021

    Bao, S., He, H., Wang, F., Wu, H., Wang, H., Wu, W., Guo, Z., Liu, Z., Xu, X.: PLATO-2: towards building an open-domain chatbot via curriculum learning. In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021. Findings of ACL, vol. ACL/IJCNLP 2021, pp. 2513–2525 (2021)

    work page 2021

  4. [4]

    In: ICML 2009

    Bengio, Y., Louradour, J., Collobert, R., Weston, J.: Curriculum learning. In: ICML 2009. ACM International Conference Proceeding Series, vol. 382, pp. 41–48 (2009)

    work page 2009

  5. [5]

    Advances in neural information processing systems 33, 1877–1901 (2020)

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Nee- lakantan, A., Shyam, P., Sastry, G., Askell, A., et al.: Language models are few-shot learners. Advances in neural information processing systems 33, 1877–1901 (2020)

    2020

  6. [6]

    In: WWW 2022

    Chen, X., Zhang, N., Xie, X., Deng, S., Yao, Y., Tan, C., Huang, F., Si, L., Chen, H.: Knowprompt: Knowledge-aware prompt-tuning with synergistic optimization for relation extraction. In: WWW 2022. pp. 2778–2788 (2022) CPTuning: Contrastive Prompt Tuning for Generative Relation Extraction 15

    work page 2022

  7. [7]

    In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021

    Cui, L., Wu, Y., Liu, J., Yang, S., Zhang, Y.: Template-based named entity recog- nition using BART. In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021. Findings of ACL, vol. ACL/IJCNLP 2021, pp. 1835–1845 (2021)

    work page 2021

  8. [8]

    Detroja, K., Bhensdadia, C.K., Bhatt, B.S.: A survey on relation extraction. Intell. Syst. Appl. 19, 200244 (2023)

    work page 2023

Show all 63 references
  1. [9]

    In: NAACL-HLT 2019

    Devlin, J., Chang, M., Lee, K., Toutanova, K.: BERT: pre-training of deep bidirec- tional transformers for language understanding. In: NAACL-HLT 2019. pp. 4171– 4186 (2019)

    2019

  2. [10]

    Guo, J., Tan, X., Xu, L., Qin, T., Chen, E., Liu, T.: Fine-tuning by curriculum learning for non-autoregressive neural machine translation. In: The Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2020, The Thirty-Second In- novative Applications of Artificial In...

    2020

  3. [11]

    In: Findings of the Association for Computational Linguistics: EMNLP 2022

    Han, J., Zhao, S., Cheng, B., Ma, S., Lu, W.: Generative prompt tuning for rela- tion classification. In: Findings of the Association for Computational Linguistics: EMNLP 2022. pp. 3170–3185 (2022)

    2022

  4. [12]

    AI Open 3, 182–192 (2022)

    Han, X., Zhao, W., Ding, N., Liu, Z., Sun, M.: PTR: prompt tuning with rules for text classification. AI Open 3, 182–192 (2022)

    2022

  5. [13]

    CoRR abs/2012.04821 (2020)

    Jiang, H., Bao, Q., Cheng, Q., Yang, D., Wang, L., Xiao, Y.: Complex relation extraction: Challenges and opportunities. CoRR abs/2012.04821 (2020)

    2020 arXiv

  6. [14]

    Joshi, M., Chen, D., Liu, Y., Weld, D.S., Zettlemoyer, L., Levy, O.: Spanbert: Im- proving pre-training by representing and predicting spans. Trans. Assoc. Comput. Linguistics 8, 64–77 (2020)

    2020

  7. [15]

    In: Proceedings of the 2022 Conference of the North Amer- ican Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL 2022

    Josifoski, M., Cao, N.D., Peyrard, M., Petroni, F., West, R.: Genie: Generative information extraction. In: Proceedings of the 2022 Conference of the North Amer- ican Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL 2022. pp. 4626–46...

    2022

  8. [16]

    In: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, EMNLP 2021

    Lester, B., Al-Rfou, R., Constant, N.: The power of scale for parameter-efficient prompt tuning. In: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, EMNLP 2021. pp. 3045–3059 (2021)

    2021

  9. [17]

    In: CoNLL 2017

    Levy, O., Seo, M., Choi, E., Zettlemoyer, L.: Zero-shot relation extraction via reading comprehension. In: CoNLL 2017. pp. 333–342 (2017)

    2017

  10. [18]

    In: ACL 2020

    Lewis, M., Liu, Y., Goyal, N., Ghazvininejad, M., Mohamed, A., Levy, O., Stoy- anov, V., Zettlemoyer, L.: BART: denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In: ACL 2020. pp. 7871–7880 (2020)

    2020

  11. [19]

    Li, X.L., Liang, P.: Prefix-tuning: Optimizing continuous prompts for generation. In: Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Pro- cessing, ACL/IJCNLP 2021, (Volume ...

    2021

  12. [20]

    In: ACL 2019

    Li, X., Yin, F., Sun, Z., Li, X., Yuan, A., Chai, D., Zhou, M., Li, J.: Entity-relation extraction as multi-turn question answering. In: ACL 2019. pp. 1340–1350 (2019)

    2019

  13. [21]

    In: IJCAI 2020

    Liu, J., Ren, Y., Tan, X., Zhang, C., Qin, T., Zhao, Z., Liu, T.: Task-level curricu- lum learning for non-autoregressive neural machine translation. In: IJCAI 2020. pp. 3861–3867 (2020)

    2020

  14. [22]

    ACM Comput

    Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., Neubig, G.: Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Comput. Surv. 55(9), 195:1–195:35 (2023) 16 Duan et al

    2023

  15. [23]

    CoRR abs/2103.10385 (2021)

    Liu, X., Zheng, Y., Du, Z., Ding, M., Qian, Y., Yang, Z., Tang, J.: GPT under- stands, too. CoRR abs/2103.10385 (2021)

    2021 arXiv

  16. [24]

    In: ACL 2020

    Liu, X., Lai, H., Wong, D.F., Chao, L.S.: Norm-based curriculum learning for neural machine translation. In: ACL 2020. pp. 427–436 (2020)

    2020

  17. [25]

    CoRR abs/1907.11692 (2019)

    Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., Stoyanov, V.: Roberta: A robustly optimized BERT pretraining approach. CoRR abs/1907.11692 (2019)

    2019 arXiv

  18. [26]

    In: Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2022

    Liu, Y., Liu, P., Radev, D.R., Neubig, G.: BRIO: bringing order to abstractive summarization. In: Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2022. pp. 2890–2903 (2022)

    2022

  19. [27]

    arXiv preprint arXiv:1711.05101 (2017)

    Loshchilov, I., Hutter, F.: Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101 (2017)

    2017 arXiv

  20. [28]

    In: Findings of the Association for Computational Linguistics: EMNLP 2022

    Lu, K., Hsu, I., Zhou, W., Ma, M.D., Chen, M.: Summarization as indirect super- vision for relation extraction. In: Findings of the Association for Computational Linguistics: EMNLP 2022. pp. 6575–6594 (2022)

    2022

  21. [29]

    In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021

    Lyu, S., Chen, H.: Relation classification with entity type restriction. In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021. Findings of ACL, vol. ACL/IJCNLP 2021, pp. 390–395 (2021)

    2021

  22. [30]

    In: Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies

    Maharana, A., Bansal, M.: On curriculum learning for commonsense reasoning. In: Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. pp. 983–992 (2022)

    2022

  23. [31]

    M¨ uller, R., Kornblith, S., Hinton, G.E.: When does label smoothing help? In: Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019. pp. 4696–4705 (2019)

    2019

  24. [32]

    In: Proceedings of the first workshop on fact extraction and VERification (FEVER)

    Obamuyide, A., Vlachos, A.: Zero-shot relation classification as textual entail- ment. In: Proceedings of the first workshop on fact extraction and VERification (FEVER). pp. 72–78 (2018)

    2018

  25. [33]

    In: ICLR 2021 (2021)

    Paolini, G., Athiwaratkun, B., Krone, J., Ma, J., Achille, A., Anubhai, R., dos Santos, C.N., Xiang, B., Soatto, S.: Structured prediction as translation between augmented natural languages. In: ICLR 2021 (2021)

    2021

  26. [34]

    arXiv preprint arXiv:2107.09332 (2021)

    Park, S., Kim, H.: Improving sentence-level relation extraction through curriculum learning. arXiv preprint arXiv:2107.09332 (2021)

    2021 arXiv

  27. [35]

    In: EMNLP

    Peng, H., Gao, T., Han, X., Lin, Y., Li, P., Liu, Z., Sun, M., Zhou, J.: Learning from context or names? an empirical study on neural relation extraction. In: EMNLP

  28. [36]

    In: EMNLP-IJCNLP

    Peters, M.E., Neumann, M., IV, R.L.L., Schwartz, R., Joshi, V., Singh, S., Smith, N.A.: Knowledge enhanced contextual word representations. In: EMNLP-IJCNLP

  29. [37]

    In: NAACL-HLT 2019

    Platanios, E.A., Stretcu, O., Neubig, G., P´ oczos, B., Mitchell, T.M.: Competence- based curriculum learning for neural machine translation. In: NAACL-HLT 2019. pp. 1162–1172 (2019)

    2019

  30. [38]

    Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., Liu, P.J.: Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res. 21, 140:1–140:67 (2020)

    2020

  31. [39]

    In: Balc´ azar, J.L., Bonchi, F., Gionis, A., Sebag, M

    Riedel, S., Yao, L., McCallum, A.: Modeling relations and their mentions without labeled text. In: Balc´ azar, J.L., Bonchi, F., Gionis, A., Sebag, M. (eds.) Machine Learning and Knowledge Discovery in Databases, European Conference, ECML PKDD 2010. Lecture Notes in Computer S...

    2010

  32. [40]

    arXiv preprint arXiv:2109.03659 (2021)

    Sainz, O., de Lacalle, O.L., Labaka, G., Barrena, A., Agirre, E.: Label verbalization and entailment for effective zero-and few-shot relation extraction. arXiv preprint arXiv:2109.03659 (2021)

    2021 arXiv

  33. [41]

    In: ACL 2019

    Soares, L.B., FitzGerald, N., Ling, J., Kwiatkowski, T.: Matching the blanks: Dis- tributional similarity for relation learning. In: ACL 2019. pp. 2895–2905 (2019)

    2019

  34. [42]

    Soviany, P., Ionescu, R.T., Rota, P., Sebe, N.: Curriculum learning: A survey. Int. J. Comput. Vis. 130(6), 1526–1565 (2022)

    2022

  35. [43]

    Stoica, G., Platanios, E.A., P´ oczos, B.: Re-tacred: Addressing shortcomings of the TACRED dataset. In: Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Thirty-Third Conference on Innovative Applications of Artificial Intel- ligence, IAAI 2021, The Eleventh...

    2021

  36. [44]

    CoRR abs/1610.02424 (2016)

    Vijayakumar, A.K., Cogswell, M., Selvaraju, R.R., Sun, Q., Lee, S., Crandall, D.J., Batra, D.: Diverse beam search: Decoding diverse solutions from neural sequence models. CoRR abs/1610.02424 (2016)

    2016 arXiv

  37. [45]

    In: EMNLP 2021

    Wang, C., Liu, X., Chen, Z., Hong, H., Tang, J., Song, D.: Zero-shot information extraction as a unified text-to-triple translation. In: EMNLP 2021. pp. 1225–1238 (2021)

    2021

  38. [46]

    In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021

    Wang, R., Tang, D., Duan, N., Wei, Z., Huang, X., Ji, J., Cao, G., Jiang, D., Zhou, M.: K-adapter: Infusing knowledge into pre-trained models with adapters. In: Findings of the Association for Computational Linguistics: ACL/IJCNLP 2021. Findings of ACL, vol. ACL/IJCNLP 2021, p...

    2021

  39. [47]

    IEEE Trans

    Wang, X., Chen, Y., Zhu, W.: A survey on curriculum learning. IEEE Trans. Pattern Anal. Mach. Intell. 44(9), 4555–4576 (2022)

    2022

  40. [48]

    In: EMNLP 2020

    Wolf, T., Debut, L., Sanh, V., Chaumond, J., Delangue, C., Moi, A., Cistac, P., Rault, T., Louf, R., Funtowicz, M., Davison, J., Shleifer, S., von Platen, P., Ma, C., Jernite, Y., Plu, J., Xu, C., Scao, T.L., Gugger, S., Drame, M., Lhoest, Q., Rush, A.M.: Transformers: State-o...

    2020

  41. [49]

    In: Proceedings of the 28th ACM International Conference on Information and Knowledge Management, CIKM 2019

    Wu, S., He, Y.: Enriching pre-trained language model with entity information for relation classification. In: Proceedings of the 28th ACM International Conference on Information and Knowledge Management, CIKM 2019. pp. 2361–2364 (2019)

    2019

  42. [50]

    In: ACL 2020

    Xu, B., Zhang, L., Mao, Z., Wang, Q., Xie, H., Zhang, Y.: Curriculum learning for natural language understanding. In: ACL 2020. pp. 6095–6104 (2020)

    2020

  43. [51]

    Xue, F., Sun, A., Zhang, H., Chng, E.S.: Gdpnet: Refining latent multi-view graph for relation extraction. In: Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Thirty-Third Conference on Innovative Applications of Artificial Intel- ligence, IAAI 2021, The El...

    2021

  44. [52]

    In: EMNLP 2020

    Yamada, I., Asai, A., Shindo, H., Takeda, H., Matsumoto, Y.: LUKE: deep contex- tualized entity representations with entity-aware self-attention. In: EMNLP 2020. pp. 6442–6454 (2020)

    2020

  45. [53]

    Yang, S., Song, D.: FPC: fine-tuning with prompt curriculum for relation extrac- tion. In: Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Con- ference on Natural Language Processin...

    2022

  46. [54]

    In: NAACL- HLT 2021

    Yasunaga, M., Ren, H., Bosselut, A., Liang, P., Leskovec, J.: QA-GNN: reasoning with language models and knowledge graphs for question answering. In: NAACL- HLT 2021. pp. 535–546 (2021) 18 Duan et al

    2021

  47. [55]

    In: EMNLP 2018

    Zhang, Y., Qi, P., Manning, C.D.: Graph convolution over pruned dependency trees improves relation extraction. In: EMNLP 2018. pp. 2205–2215 (2018)

    2018

  48. [56]

    In: EMNLP 2017

    Zhang, Y., Zhong, V., Chen, D., Angeli, G., Manning, C.D.: Position-aware atten- tion and supervised data improve slot filling. In: EMNLP 2017. pp. 35–45 (2017)

    2017

  49. [57]

    In: ACL 2019

    Zhang, Z., Han, X., Liu, Z., Jiang, X., Sun, M., Liu, Q.: ERNIE: enhanced language representation with informative entities. In: ACL 2019. pp. 1441–1451 (2019)

    2019

  50. [58]

    CoRR abs/2306.02051 (2023)

    Zhao, X., Deng, Y., Yang, M., Wang, L., Zhang, R., Cheng, H., Lam, W., Shen, Y., Xu, R.: A comprehensive survey on deep learning for relation extraction: Recent advances and new frontiers. CoRR abs/2306.02051 (2023)

    2023 arXiv

  51. [59]

    In: EMNLP 2020

    Zhao, X., Wu, W., Xu, C., Tao, C., Zhao, D., Yan, R.: Knowledge-grounded dia- logue generation with pre-trained language models. In: EMNLP 2020. pp. 3377– 3390 (2020)

    2020

  52. [60]

    In: EMNLP 2021

    Zhou, W., Chen, M.: Learning from noisy labels for entity-centric information extraction. In: EMNLP 2021. pp. 5381–5392 (2021)

    2021

  53. [61]

    In: AACL/IJCNLP 2022

    Zhou, W., Chen, M.: An improved baseline for sentence-level relation extraction. In: AACL/IJCNLP 2022. pp. 161–168 (2022)

    2022

  54. [62]

    In: ACL 2020

    Zhou, Y., Yang, B., Wong, D.F., Wan, Y., Chao, L.S.: Uncertainty-aware curricu- lum learning for neural machine translation. In: ACL 2020. pp. 6934–6944 (2020)

    2020

  55. [2020]

    3661–3672 (2020)

    pp. 3661–3672 (2020)

    2020

Pith tools

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