Probing Neural Network Comprehension of Natural Language Arguments
Open accessThe DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index.
Abstract
We are surprised to find that BERT's peak performance of 77% on the Argument Reasoning Comprehension Task reaches just three points below the average untrained human baseline.However, we show that this result is entirely accounted for by exploitation of spurious statistical cues in the dataset.We analyze the nature of these cues and demonstrate that a range of models all exploit them.This analysis informs the construction of an adversarial dataset on which all models achieve random accuracy.Our adversarial dataset provides a more robust assessment of argument comprehension and should be adopted as the standard in future work.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Figures


Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“This phenomenon is known as shortcut learning31 and underlies many deep-learning failures32.”
From A foundation model for clinician-centered drug repurposing (2024) · cites this paper as [32]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 4×
“The BoV and BiLSTM inputs are 300 -dimensional GloVe embeddings trained on 640 B tokens (Pennington et al., 2014).”
MethodInfluentialAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 1×
“For all of our experiments we use grid search to select hyperparameters, dropout regularization (Sri-vastava et al., 2014), and Adam (Kingma and Ba, 2014) for optimization.”
MethodBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173 · cited 3×
“3 Devlin et al. (2018) report that, on small datasets, BERT sometimes fails to train, yielding degenerate results.”
BackgroundInfluentialBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325 · cited 3×
“3 Devlin et al. (2018) report that, on small datasets, BERT sometimes fails to train, yielding degenerate results.”
BackgroundInfluentialGlove: Global Vectors for Word Representation
Pennington, J., Socher, R., & Manning, C. (2014) · Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1162 · Cited by 34,071 · cited 1×
“The BoV and BiLSTM inputs are 300 -dimensional GloVe embeddings trained on 640 B tokens (Pennington et al., 2014).”
Background(2014) · Cited by 33,419
Toulmin, S. E. (2003) · doi:10.1017/cbo9780511840005 · Cited by 7,101
A large annotated corpus for learning natural language inference
Bowman, S. R., Angeli, G., Potts, C., & Manning, C. D. (2015) · Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d15-1075 · Cited by 3,377 · cited 1×
“…work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and InferSent (Conneau et al., 2017).”
MethodA Broad-Coverage Challenge Corpus for Sentence Understanding through Inference
Williams, A., Nangia, N., & Bowman, S. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) · doi:10.18653/v1/n18-1101 · Cited by 3,342 · cited 1×
“…work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and InferSent (Conneau et al., 2017).”
MethodKnow What You Don't Know: Unanswerable Questions for SQuAD
Rajpurkar, P., Jia, R., & Liang, P. (2018) · arXiv · doi:10.48550/arxiv.1806.03822 · Cited by 2,242 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundKnow What You Don’t Know: Unanswerable Questions for SQuAD
Rajpurkar, P., Jia, R., & Liang, P. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) · doi:10.18653/v1/p18-2124 · Cited by 2,240 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundSupervised Learning of Universal Sentence Representations from Natural Language Inference Data
Conneau, A., Kiela, D., Schwenk, H., Barrault, L., & Bordes, A. (2017) · Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d17-1070 · Cited by 2,049 · cited 2×
“…work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and InferSent (Conneau et al., 2017).”
MethodAdversarial Examples for Evaluating Reading Comprehension Systems
Jia, R., & Liang, P. (2017) · Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d17-1215 · Cited by 1,410 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundRight for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference
McCoy, R. T., Pavlick, E., & Linzen, T. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1334 · Cited by 944 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundAnnotation Artifacts in Natural Language Inference Data
Gururangan, S., Swayamdipta, S., Levy, O., Schwartz, R., Bowman, S., & Smith, N. A. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers) · doi:10.18653/v1/n18-2017 · Cited by 907 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundHypothesis Only Baselines in Natural Language Inference
Poliak, A., Naradowsky, J., Haldar, A., Rudinger, R., & Van Durme, B. (2018) · Proceedings of the Seventh Joint Conference on Lexical and Computational Semantics · doi:10.18653/v1/s18-2023 · Cited by 480 · cited 2×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundResultA large annotated corpus for learning natural language inference
Bowman, S. R., Angeli, G., Potts, C. E., & Manning, C. D. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1508.05326 · Cited by 381 · cited 1×
“…work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and InferSent (Conneau et al., 2017).”
MethodInformal logic: a handbook for critical argumentation
(1990) · Choice Reviews Online · doi:10.5860/choice.27-4477 · Cited by 374
Argumentation Mining: State of the Art and Emerging Trends
Lippi, M., & Torroni, P. (2016) · ACM Transactions on Internet Technology · doi:10.1145/2850417 · Cited by 363
Breaking NLI Systems with Sentences that Require Simple Lexical Inferences
Glockner, M., Shwartz, V., & Goldberg, Y. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) · doi:10.18653/v1/p18-2103 · Cited by 319 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundAdversarial Examples for Evaluating Reading Comprehension SystemsPreprint
Jia, R., & Liang, P. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1707.07328 · Cited by 259 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundMochales, R., & Moens, M.-F. (2011) · Artificial Intelligence and Law · doi:10.1007/s10506-010-9104-x · Cited by 252 · cited 1×
“Argumentation mining is the task of determining argumentative structure in natural language text - e.g., which text segments represent claims, and which comprise reasons that support or attack those claims (Mochales and Moens, 2011; Lippi and Torroni, 2016).”
BackgroundMeasuring the tendency of CNNs to Learn Surface Statistical Regularities
Jo, J., & Bengio, Y. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1711.11561 · Cited by 173 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundThe Argument Reasoning Comprehension Task: Identification and Reconstruction of Implicit Warrants
Habernal, I., Wachsmuth, H., Gurevych, I., & Stein, B. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) · doi:10.18653/v1/n18-1175 · Cited by 121 · cited 14×
“The most successful previous work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and InferSent (Conneau et al.…”
BackgroundMethodResultInfluentialArgumentation Mining on the Web from Information Seeking Perspective
Habernal, I., Eckle‐Kohler, J., & Gurevych, I. (2014) · TUbilio (Technical University of Darmstadt) · Cited by 94
Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language InferencePreprint
McCoy, R. T., Pavlick, E., & Linzen, T. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1902.01007 · Cited by 86 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundEnhancing and Combining Sequential and Tree LSTM for Natural Language Inference.Preprint
Qian, C., Zhu, X., Ling, Z.-H., Wei, S., & Jiang, H. (2016) · arXiv (Cornell University) · Cited by 82
Behavior Analysis of NLI Models: Uncovering the Influence of Three Factors on Robustness
Sanchez, I., Mitchell, J., & Riedel, S. (2018) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/n18-1179 · Cited by 55 · cited 2×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundA Relevance-theoretic Framework for Constructing and Deconstructing Enthymemes
Black, E., & Anthony Hunter, A. (2009) · Journal of Logic and Computation · doi:10.1093/logcom/exp064 · Cited by 35 · cited 1×
“This example adapted from Black and Hunter (2012) Claim Google is not a harmful monopoly Reason People can choose not to use Google Warrant Other search engines don’t redirect to Google Alternative All other search engines redirect to Google”
SemEval-2018 Task 12: The Argument Reasoning Comprehension Task
Habernal, I., Wachsmuth, H., Gurevych, I., & Stein, B. (2018) · International Workshop on Semantic Evaluation (SemEval) · doi:10.18653/v1/s18-1121 · Cited by 24 · cited 13×
“The most successful previous work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and InferSent (Conneau et al.…”
BackgroundMethodResultInfluentialFrame- and Entity-Based Knowledge for Common-Sense Argumentative Reasoning
Botschen, T., Sorokin, D., & Gurevych, I. (2018) · Workshop on Argument Mining (ArgMining) · doi:10.18653/v1/w18-5211 · Cited by 23 · cited 2×
“The baselines are a bag of vectors (BoV), bidirectional LSTM (Hochreiter and Schmidhu-ber, 1997) (BiLSTM), the SemEval winner GIST (Choi and Lee, 2018), the best model of Botschen et al. (2018), and human performance (Table 1).”
BackgroundMethodChoi, H., & Lee, H. (2018) · International Workshop on Semantic Evaluation (SemEval) · doi:10.18653/v1/s18-1122 · Cited by 15 · cited 2×
“The most successful previous work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and…”
MethodBreaking NLI Systems with Sentences that Require Simple Lexical InferencesPreprint
Glockner, M., Shwartz, V., & Goldberg, Y. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1805.02266 · Cited by 8 · cited 1×
“There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
BackgroundNLITrans at SemEval-2018 Task 12: Transfer of Semantic Knowledge for Argument Comprehension
Niven, T., & Kao, H. (2018) · International Workshop on Semantic Evaluation (SemEval) · doi:10.18653/v1/s18-1185 · Cited by 5 · cited 3×
“The most successful previous work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and…”
BackgroundMethodSelf-citationZhao, M., Liu, C., Liu, L., Yan, Z., & Yu, D. (2018) · International Workshop on Semantic Evaluation (SemEval) · doi:10.18653/v1/s18-1186 · Cited by 3 · cited 13×
“The most successful previous work on ARCT (Choi and Lee, 2018; Zhao et al., 2018; Niven and Kao, 2018) involved transfer learning from Natural Language Inference (NLI) datasets (Bowman et al., 2015; Williams et al., 2017), and utilized effective NLI models such as ESIM (Chen et al., 2016) and InferSent (Conneau et al.…”
BackgroundInfluential
Research works printed but not linked: 1
- [27]27. Douglas N. Walton. 2005. Informal logic: a hand- book of critical argumentation. Cambridge Univer- sity Press.
Cited by
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
“The main disadvantages are the need for a new large dataset for every task, the potential for poor generalization out-of-distribution [MPL19], and the potential to exploit spurious features of the training data [GSL + 18, NK19].”
BackgroundOn the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445922 · Cited by 6,988
“However, no actual language understanding is taking place in LM-driven approaches to these tasks, as can be shown by careful manipulation of the test data to remove spurious cues the systems are leveraging [21, 93].”
BackgroundLiu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., & Neubig, G. (2023) · ACM Computing Surveys · doi:10.1145/3560815 · Cited by 4,361
“5 łReasoningž in NLP There is still a debate(7) about if deep neural networks are capable of performing łreasoningž or just memorizing patterns based on large training data [3, 89].”
BackgroundShortcut learning in deep neural networks
Geirhos, R., Jacobsen, J.-H., Michaelis, C., Zemel, R., Brendel, W., Bethge, M., & Wichmann, F. A. (2020) · Nature Machine Intelligence · doi:10.1038/s42256-020-00257-z · Cited by 2,224
“Attempts like incorporating a certain degree of unsupervised training as employed in prominent language models like BERT [5] and GPT-2 [66] did not resolve the problem of shortcut learning [14].”
BackgroundInfluentialA primer in BERTology: What we know about how BERT works
Rogers, A., Kovaleva, O., & Rumshisky, A. (2020) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00349 · Cited by 1,552
Metadata
- Authors
- Niven, T., & Kao, H.-Y.
- Affiliations
- National Cheng Kung University (TW)
- Corresponding author
- Kao, Hung-Yu
- ORCID
- Kao, Hung-Yu
- Venue
- Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics
- Published
- 2019
- Pages
- 4658-4664
- ISSN
- 0736-587X (ISSN-L)
- OpenAlex
- W2952984539
- Cited by
- 395
- References
- 31
- Citations by year
- 2026: 9 · 2025: 18 · 2024: 35 · 2023: 63 · 2022: 46 · 2021: 103 · 2020: 97 · 2019: 23
- FWCI
- 25.82
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Proceedings of the conference - Association for Computational Linguistics. Meeting (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Keywords
- argument comprehension, BERT, data bias, adversarial datasets, model robustness
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.