---
title: "Predicting the Argumenthood of English Prepositional Phrases"
description: "The authors define two tasks. 95.5% accuracy. 38 references, cited by 1 paper here. Among them: 3."
authors: ["Najoung Kim", "Kyle Rawlins", "Benjamin Van Durme", "Paul Smolensky"]
year: 2018
venue: "arXiv:1809.07889"
source: "https://arxiv.org/abs/1809.07889"
doi: "10.48550/arxiv.1809.07889"
arxiv: "1809.07889"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/kim-2018-predicting-argumenthood-english-prepositional-phrases/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/kim-2018-predicting-argumenthood-english-prepositional-phrases/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2018](https://md.research.elamur.ai/free/en/papers/2018/page/6/) / Predicting the Argumenthood of English Prepositional Phrases

# Predicting the Argumenthood of English Prepositional Phrases

**Authors:** [Najoung Kim](https://md.research.elamur.ai/free/en/authors/najoung-kim/), [Kyle Rawlins](https://md.research.elamur.ai/free/en/authors/kyle-rawlins/), [Benjamin Van Durme](https://md.research.elamur.ai/free/en/authors/benjamin-van-durme/) and [Paul Smolensky](https://md.research.elamur.ai/free/en/authors/paul-smolensky/) · **Year:** 2018 · **Venue:** [arXiv:1809.07889](https://md.research.elamur.ai/free/en/venues/proceedings-of-the-aaai-conference-on-artificial-intelligence/) ([Proceedings of the AAAI Conference on Artificial Intelligence 2018](https://md.research.elamur.ai/free/en/venues/proceedings-of-the-aaai-conference-on-artificial-intelligence/2018/)) · **Publisher:** [Association for the Advancement of Artificial Intelligence](https://md.research.elamur.ai/free/en/publishers/association-for-the-advancement-of-artificial-intelligence/) · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/1809.07889) · [PDF](https://arxiv.org/pdf/1809.07889) · [DOI 10.48550/arxiv.1809.07889](https://doi.org/10.48550/arxiv.1809.07889) · [arXiv 1809.07889](https://arxiv.org/abs/1809.07889) · [OpenAlex](https://openalex.org/W2966638880)

**Citations:** 3 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 141,840 papers whose bibliographies we read

## Summary
The paper asks whether a model can tell if a prepositional phrase attached to a verb is an argument, meaning it is selected by the verb, or an adjunct, meaning an optional modifier. Linguistic diagnostics for this distinction often disagree, and major treebanks leave it unannotated. The authors define two tasks. The first is binary classification of verb–preposition pairs, with labels taken from VerbNet subcategorization frames. The second is regression onto graded human judgments collected by crowdsourcing. Their models combine pre-trained word embeddings (word2vec, GloVe, ELMo and others) with linguistically motivated features. On the balanced binary task, an ELMo+BiLSTM model reaches about 95.5% accuracy. On the gradient task, word2vec+MLP reaches Pearson r = 0.624. They also pretrain a sentence encoder on these tasks, freeze it, and train SRL classifiers on top. This beats a randomly initialized encoder on both CoNLL SRL benchmarks. It also beats an encoder pretrained directly on PP attachment. Argumenthood therefore appears to be a useful standalone signal for parsing and role labeling.

## Key points
- The candidate space covers 2714 VerbNet verbs and 60 single-word prepositions, giving 162,480 {V, P} pairs. A pair is labeled ARG if VerbNet lists that PP frame for the verb and ADJ otherwise. Feature-only frames such as {+SPATIAL} were mapped to preposition sets by hand, using PrepWiki senses.
- Arguments are rare, at about 1:10 against adjuncts. The authors kept all 13,544 positives and an equal random sample of negatives, giving n = 27,088. This was split 70:15:15, so the test set has n = 4064. The best result is ELMo+BiLSTM with 95.5 accuracy and 95.4 F1.
- Full-sentence variants, including a ternary one, were built by searching the SNLI and MNLI parse trees for sentences containing each {V, P}. These variants are noisier because the trees are parser output and the labels stay type-level.
- Frozen-encoder pretraining used the jiant setup of Bowman et al. 2018. On SRL-CoNLL2005 F1 rose from 81.7 (random) to 83.9–84.7. On SRL-CoNLL2012 it rose from 77.3 to 80.2–80.7, with p \< .001 by Approximate Randomization. There was no significant gain on the Belinkov et al. 2014 PP attachment task: Arg fullsent scored 88.2, against 88.7 for an encoder trained on attachment directly.
- An encoder pretrained on PP attachment did worse on SRL (80.2 / 79.8) than any argumenthood-pretrained encoder. The authors also show that dataset size alone does not explain the gains. A new public dataset of gradient judgments was released.

## Interesting ideas
- Treat membership in a curated lexicon, here VerbNet frames, as a cheap type-level gold label for a hard-to-annotate property, then rebalance by subsampling the dominant class.
- Turn type-level labels into token-level training examples by searching parsed corpora for sentences that contain the construction, while openly treating the result as noisier data.
- Model argumenthood as a graded score fitted to human ratings rather than a hard yes/no, and treat the leftover disagreement as informative.
- Probe what an auxiliary task teaches by freezing an encoder pretrained on it and comparing against a random-init encoder. Add two controls: a competing related task (PP attachment) and differing dataset sizes. Test significance with approximate randomization.
- The paper notes that UD mixes the core/non-core split with the argument/adjunct split, while grammars like CCG encode argumenthood explicitly, so a predictor helps when porting resources between formalisms.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (frames): Mova already draws on VerbNet, so a VerbNet-derived ARG/ADJ prior per verb–preposition pair could decide whether a PP fills a role slot or becomes a modifier, and the decision can be explained by pointing at the listed frame.
- **I2** (learn): silver sentences for verb+PP constructions could be mined this way and kept only where several witness parsers agree on the attachment, which reduces the parser-noise problem the authors point out.
- **I3** (frames): storing a graded argumenthood score lets the frame builder hedge on borderline PPs, such as instrument-like "with", and report how confident it is instead of forcing a brittle choice.
- **I4** (learn): this is a ready template for negative controls when checking whether a new training signal helps parsing or role labeling, rather than just adding more data.
- **I5** (tree): an argumenthood predictor could inform obl-vs-argument decisions and PP attachment features in the perceptron parser and gparse, and help when converting between treebank dialects.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Johns Hopkins University](https://md.research.elamur.ai/free/en/universities/johns-hopkins-university/) — University, Baltimore, United States · 187 works · 92,194 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 1,373 works

## Cite

Cite as: Najoung Kim, Kyle Rawlins, Benjamin Van Durme and Paul Smolensky. 2018. Predicting the Argumenthood of English Prepositional Phrases. arXiv:1809.07889.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{kim2018predicting,
  title = {{Predicting the Argumenthood of English Prepositional Phrases}},
  author = {Najoung Kim, Kyle Rawlins, Benjamin {Van Durme} and Paul Smolensky},
  year = {2018},
  doi = {10.48550/arxiv.1809.07889},
  eprint = {1809.07889},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1809.07889}
}
```

- [← previous Neural Surface Realization for Italian](https://md.research.elamur.ai/free/en/papers/basile-2018-neural-surface-realization-italian/)
- [next → SentiArabic: A Sentiment Analyzer for Standard Arabic](https://md.research.elamur.ai/free/en/papers/eskander-2018-sentiarabic-sentiment-analyzer-standard-arabic/)
- [↑ back to papers of 2018](https://md.research.elamur.ai/free/en/papers/2018/page/6/)

## References

38 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. [2010] Abend, O., and Rappoport, A. 2010. Fully unsupervised core-adjunct argument classification. In Proc. of the 48th Annual Meeting of the Association for Computational Linguistics, 226– 236. [find](https://scholar.google.com/scholar?q=%5B2010%5D+Abend%2C+O%2E%2C+and+Rappoport%2C+A%2E+2010%2E+Fully+unsupervised+core%2Dadjunct+argument+classification%2E+In+Proc%2E+of+the+48th+Annual+Meeting+of+the+Association+for+Computational+Linguistics%2C+226%E2%80%93+236%2E)
2. [2008] Agirre, E.; Baldwin, T.; and Martinez, D. 2008. Improving parsing and PP attachment performance with sense information. In Proc. of the 46th Annual Meeting of the Association for Computational Linguistics, 317–325. [find](https://scholar.google.com/scholar?q=%5B2008%5D+Agirre%2C+E%2E%3B+Baldwin%2C+T%2E%3B+and+Martinez%2C+D%2E+2008%2E+Improving+parsing+and+PP+attachment+performance+with+sense+information%2E+In+Proc%2E+of+the+46th+Annual+Meeting+of+the+Association+for+Computational+)
3. [2002] Aldezabal, I.; Aranzabe, M.; Gojenola, K.; Sarasola, K.; and Atutxa, A. 2002. Learning argument/adjunct distinction for To be released at: decomp.io basque. In Proc. of the ACL-02 Workshop on Unsupervised lexical acquisition, Vol. 9, 42–50. [find](https://scholar.google.com/scholar?q=%5B2002%5D+Aldezabal%2C+I%2E%3B+Aranzabe%2C+M%2E%3B+Gojenola%2C+K%2E%3B+Sarasola%2C+K%2E%3B+and+Atutxa%2C+A%2E+2002%2E+Learning+argument%2Fadjunct+distinction+for+To+be+released+at%3A+decomp%2Eio+basque%2E+In+Proc%2E+of+the+ACL%2D02+Workshop+on+U)
4. [2014] Belinkov, Y.; Lei, T.; Barzilay, R.; and Globerson, A. 2014. [Exploring compositional architectures and word vector representations for prepositional phrase attachment](https://md.research.elamur.ai/free/en/papers/belinkov-2014-exploring-compositional-architectures-word-vector/). Transactions of the Association for Computational Linguistics 2:561–572. [↗](https://doi.org/10.1162/tacl_a_00203)
5. [2016] Bojanowski, P.; Grave, E.; Joulin, A.; and Mikolov, T. 2016. [Enriching word vectors with subword information](https://md.research.elamur.ai/free/en/papers/bojanowski-2017-enriching-word-vectors-subword-information/). arXiv:1607.04606. [↗](https://arxiv.org/abs/1607.04606)
6. [2015] Bowman, S. R.; Angeli, G.; Potts, C.; and Manning, C. D. 2015. A large annotated corpus for learning natural language inference. In Proc. of the 2015 Conference on Empirical Methods in Natural Language Processing, 632–642. [↗](https://doi.org/10.48550/arxiv.1508.05326)
7. [2018] Bowman, S. R.; Pavlick, E.; Grave, E.; Durme, B. V.; Wang, A.; Hula, J.; Xia, P.; Pappagari, R.; McCoy, R. T.; Patel, R.; Kim, N.; Tenney, I.; Huang, Y.; Yu, K.; Jin, S.; and Chen, B. 2018. Looking for elmo’s friends: Sentence-level pretraining beyond language modeling. arXiv:1812.10860. [↗](https://arxiv.org/abs/1812.10860)
8. [2005] Carreras, X., and Màrquez, L. 2005. [Introduction to the conll-2005 shared task: Semantic role labeling](https://md.research.elamur.ai/free/en/papers/carreras-2005-introduction-conll-2005-shared-task/). In Proc. of the 9th Conference on Computational Natural Language Learning, 152– 164.
9. [1998] Carroll, J.; Minnen, G.; and Briscoe, T. 1998. Can subcategorisation probabilities help a statistical parser? In Proc. of the 6th ACL/SIGDAT Workshop in Very Large Corpora. http://www.aclweb.org/anthology/W98-1114. [↗](https://aclanthology.org/W98-1114/)
10. [1993] Chomsky, N. 1993. Lectures on government and binding: The Pisa lectures. Walter de Gruyter. [find](https://scholar.google.com/scholar?q=%5B1993%5D+Chomsky%2C+N%2E+1993%2E+Lectures+on+government+and+binding%3A+The+Pisa+lectures%2E+Walter+de+Gruyter%2E)
11. [2017] Dasigi, P.; Ammar, W.; Dyer, C.; and Hovy, E. 2017. [Ontology-aware token embeddings for prepositional phrase attachment](https://md.research.elamur.ai/free/en/papers/dasigi-2017-ontology-aware-token-embeddings-prepositional/). In Proc. of the 55th Annual Meeting of the Association for Computational Linguistics (Vol 1: Long Papers), 2089–2098.
12. [2018] Ettinger, A.; Elgohary, A.; Phillips, C.; and Resnik, P. 2018. Assessing composition in sentence vector representations. In Proc. of the 27th International Conference on Computational Linguistics, 1790–1801. [find](https://scholar.google.com/scholar?q=%5B2018%5D+Ettinger%2C+A%2E%3B+Elgohary%2C+A%2E%3B+Phillips%2C+C%2E%3B+and+Resnik%2C+P%2E+2018%2E+Assessing+composition+in+sentence+vector+representations%2E+In+Proc%2E+of+the+27th+International+Conference+on+Computational+Linguisti)
13. [2018] Gardner, M.; Grus, J.; Neumann, M.; Tafjord, O.; Dasigi, P.; Liu, N. F.; Peters, M.; Schmitz, M.; and Zettlemoyer, L. 2018. [Allennlp: A deep semantic natural language processing platform](https://md.research.elamur.ai/free/en/papers/gardner-2018-allennlp-deep-semantic-natural-language/). In Proceedings of Workshop for NLP Open Source Software (NLP-OSS), 1–6.
14. [2018] Gulordava, K.; Bojanowski, P.; Grave, E.; Linzen, T.; and Baroni, M. 2018. [Colorless green recurrent networks dream hierarchically](https://md.research.elamur.ai/free/en/papers/gulordava-2018-colorless-green-recurrent-networks-dream/). In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), volume 1, 1195–1205.
15. [2017[ He, L.; Lee, K.; Lewis, M.; and Zettlemoyer, L](https://md.research.elamur.ai/free/en/papers/he-2017-deep-semantic-role-labeling-what/). 2017. Deep semantic role labeling: What works and whats next. In Proc. of the 55th Annual Meeting of the Association for Computational Linguistics (Vol. 1: Long Papers), 473–483.
16. [2002] Hockenmaier, J., and Steedman, M. 2002. Acquiring compact lexicalized grammars from a cleaner treebank. In Proc. of the 3rd International Conference on Language Resources and Evaluation. [find](https://scholar.google.com/scholar?q=%5B2002%5D+Hockenmaier%2C+J%2E%2C+and+Steedman%2C+M%2E+2002%2E+Acquiring+compact+lexicalized+grammars+from+a+cleaner+treebank%2E+In+Proc%2E+of+the+3rd+International+Conference+on+Language+Resources+and+Evaluation%2E)
17. [2005] Kipper-Schuler, K. 2005. Verbnet: A broad-coverage, comprehensive verb lexicon. Ph. D. diss., University of Pennsylvania. [↗](https://doi.org/10.5281/zenodo.8100881)
18. [2017] Marcheggiani, D., and Titov, I. 2017. [Encoding sentences with graph convolutional networks for semantic role labeling](https://md.research.elamur.ai/free/en/papers/marcheggiani-2017-encoding-sentences-graph-convolutional-networks/). In Proc. of the 2017 Conference on Empirical Methods in Natural Language Processing, 1506–1515.
19. [2008] McConville, M., and Dzikovska, M. O. 2008. Evaluating complement-modifier distinctions in a semantically annotated corpus. In Proc. of the International Conference on Language Resources and Evaluation. [find](https://scholar.google.com/scholar?q=%5B2008%5D+McConville%2C+M%2E%2C+and+Dzikovska%2C+M%2E+O%2E+2008%2E+Evaluating+complement%2Dmodifier+distinctions+in+a+semantically+annotated+corpus%2E+In+Proc%2E+of+the+International+Conference+on+Language+Resources+and+Eva)
20. [2006] Merlo, P., and Ferrer, E. E. 2006. [The notion of argument in prepositional phrase attachment](https://md.research.elamur.ai/free/en/papers/merlo-2006-notion-argument-prepositional-phrase-attachment/). Computational Linguistics 32(3):341–378. [↗](https://doi.org/10.1162/coli.2006.32.3.341)
21. [2013] Mikolov, T.; Chen, K.; Corrado, G.; and Dean, J. 2013. [Efficient estimation of word representations in vector space](https://md.research.elamur.ai/free/en/papers/mikolov-2013-efficient-estimation-word-representations-vector/). CoRR abs/1301.3781. [↗](https://arxiv.org/abs/1301.3781)
22. [2005] Palmer, M.; Gildea, D.; and Kingsbury, P. 2005. [The proposition bank: An annotated corpus of semantic roles](https://md.research.elamur.ai/free/en/papers/palmer-2005-proposition-bank-annotated-corpus-semantic/). Computational linguistics 31(1):71–106. [↗](https://doi.org/10.1162/0891201053630264)
23. [2014] Pennington, J.; Socher, R.; and Manning, C. 2014. [Glove: Global vectors for word representation](https://md.research.elamur.ai/free/en/papers/pennington-2014-glove-global-vectors-word-representation/). In Proc. of the 2014 Conference on Empirical Methods in Natural Language Processing, 1532–1543.
24. [2018] Peters, M.; Neumann, M.; Iyyer, M.; Gardner, M.; Clark, C.; Lee, K.; and Zettlemoyer, L. 2018. [Deep contextualized word representations](https://md.research.elamur.ai/free/en/papers/peters-2018-deep-contextualized-word-representations/). In Proc. of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics (Vol. 1: Long Papers), 2227–2237.
25. [1987] Pollard, C., and Sag, I. 1987. Information-Based Syntax and Semantics, Vol. 1. CSLI Publications. [find](https://scholar.google.com/scholar?q=%5B1987%5D+Pollard%2C+C%2E%2C+and+Sag%2C+I%2E+1987%2E+Information%2DBased+Syntax+and+Semantics%2C+Vol%2E+1%2E+CSLI+Publications%2E)
26. [2012] Pradhan, S.; Moschitti, A.; Xue, N.; Uryupina, O.; and Zhang, Y. 2012. [CoNLL-2012 shared task: Modeling multilingual unrestricted coreference in OntoNotes](https://md.research.elamur.ai/free/en/papers/pradhan-2012-conll-2012-shared-task-modeling/). In Joint Conference on EMNLP and CoNLL-Shared Task, 1–40.
27. [2018] Przepiórkowski, A., and Patejuk, A. 2018. Arguments and adjuncts in universal dependencies. In Proc. of the 27th International Conference on Computational Linguistics, 3837–3852. [find](https://scholar.google.com/scholar?q=%5B2018%5D+Przepio%CC%81rkowski%2C+A%2E%2C+and+Patejuk%2C+A%2E+2018%2E+Arguments+and+adjuncts+in+universal+dependencies%2E+In+Proc%2E+of+the+27th+International+Conference+on+Computational+Linguistics%2C+3837%E2%80%933852%2E)
28. [2015] Reisinger, D.; Rudinger, R.; Ferraro, F.; Harman, C.; Rawlins, K.; and Van Durme, B. 2015. [Semantic proto-roles](https://md.research.elamur.ai/free/en/papers/reisinger-2015-semantic-proto-roles/). Transactions of the Association for Computational Linguistics 3:475–488. [↗](https://doi.org/10.1162/tacl_a_00152)
29. [2015] Rissman, L.; Rawlins, K.; and Landau, B. 2015. Using instruments to understand argument structure: Evidence for gradient representation. Cognition 142:266–290. [↗](https://doi.org/10.1016/j.cognition.2015.05.015)
30. [2015] Schneider, N.; Srikumar, V.; Hwang, J. D.; and Palmer, M. 2015. [A hierarchy with, of, and for preposition supersenses](https://md.research.elamur.ai/free/en/papers/schneider-2015-hierarchy-preposition-supersenses/). In Proc. of The 9th Linguistic Annotation Workshop, 112–123.
31. [1995] Schütze, C. T. 1995. PP attachment and argumenthood. MIT working papers in linguistics 26(95):151. [find](https://scholar.google.com/scholar?q=%5B1995%5D+Schu%CC%88tze%2C+C%2E+T%2E+1995%2E+PP+attachment+and+argumenthood%2E+MIT+working+papers+in+linguistics+26%2895%29%3A151%2E)
32. [2000] Steedman, M. 2000. The syntactic process, volume 24. MIT Press. [find](https://scholar.google.com/scholar?q=%5B2000%5D+Steedman%2C+M%2E+2000%2E+The+syntactic+process%2C+volume+24%2E+MIT+Press%2E)
33. [2019] Tenney, I.; Xia, P.; Chen, B.; Wang, A.; Poliak, A.; McCoy, R. T.; Kim, N.; Durme, B. V.; Bowman, S.; Das, D.; and Pavlick, E. 2019. [What do you learn from context? probing for sentence structure in contextualized word representations](https://md.research.elamur.ai/free/en/papers/tenney-2019-learn-context-probing-sentence-structure/). In ICLR. [↗](https://doi.org/10.48550/arxiv.1905.06316)
34. [2008] Tutunjian, D., and Boland, J. E. 2008. Do we need a distinction between arguments and adjuncts? evidence from psycholinguistic studies of comprehension. Language and Linguistics Compass 2(4):631–646. [↗](https://doi.org/10.1111/j.1749-818x.2008.00071.x)
35. [2002] Villavicencio, A. 2002. Learning to distinguish pp arguments from adjuncts. In Proc. of the 6th Conference on Natural language learning, Vol. 20, 1–7. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=%5B2002%5D+Villavicencio%2C+A%2E+2002%2E+Learning+to+distinguish+pp+arguments+from+adjuncts%2E+In+Proc%2E+of+the+6th+Conference+on+Natural+language+learning%2C+Vol%2E+20%2C+1%E2%80%937%2E+Association+for+Computational+Linguistics%2E)
36. [2018] Williams, A.; Nangia, N.; and Bowman, S. 2018. A broad-coverage challenge corpus for sentence understanding through inference. In Proc. of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics (Vol. 1: Long Papers), 1112–1122. [find](https://scholar.google.com/scholar?q=%5B2018%5D+Williams%2C+A%2E%3B+Nangia%2C+N%2E%3B+and+Bowman%2C+S%2E+2018%2E+A+broad%2Dcoverage+challenge+corpus+for+sentence+understanding+through+inference%2E+In+Proc%2E+of+the+2018+Conference+of+the+North+American+Chapter+of+t)
37. [2000] Yeh, A. 2000. More accurate tests for the statistical significance of result differences. In Proc. of the 18th Conference on Computational linguistics, Vol. 2, 947–953. [↗](https://doi.org/10.48550/arxiv.cs/0008005)
38. [2012] Zeiler, M. D. 2012. [Adadelta: an adaptive learning rate method](https://md.research.elamur.ai/free/en/papers/zeiler-2012-adadelta-adaptive-learning-rate-method/). arXiv:1212.5701. [↗](https://arxiv.org/abs/1212.5701)

*Updated 24 October 2026*