Shortcut Learning of Large Language Models in Natural Language Understanding
Open accessThe DOI matches its Crossref record; 41 of the 41 references it cites were found in a registry or an index.
Abstract
Shortcuts often hinder the robustness of large language models.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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
“Recent works show that even Large-Language Models (LLMs), pre-trained on text amounting to billions of words, suffer from shortcut learning, largely due to collection artifacts in training data [7], [21].”
From Language-Based Augmentation to Address Shortcut Learning in Object-Goal Navigation (2023) · cites this paper as [21]
Citations
Cites
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 · cited 12×
“Recently, large language models (LLMs), such as BERT [8], RoBERTa [20], T5 [30], GPT-3 [4], have been reported to achieve state-of-the-art performance in a series of high-level NLU tasks.”
BackgroundInfluentialAdvances in neural information processing systems 7
Cios, K., & Shields, M. (1997) · Neurocomputing · doi:10.1016/s0925-2312(97)90018-7 · Cited by 22,059
Advances in neural information processing systems 7
(1996) · Computers & Mathematics with Applications · doi:10.1016/s0898-1221(96)90463-0 · Cited by 14,281
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., & Liu, P. J. (2019) · arXiv · doi:10.48550/arxiv.1910.10683 · Cited by 10,929 · cited 1×
“Natural language understanding (NLU) is a subfield of artificial intelligence that requires computer software to comprehend input in the form of sentences.”
BackgroundChain-Of-Thought Prompting Elicits Reasoning in Large Language Models
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q. V., & Zhou, D. (2022) · Advances in Neural Information Processing Systems 35 · doi:10.52202/068431-1800 · Cited by 4,505
Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., & Zhou, D. (2022) · arXiv · doi:10.48550/arxiv.2201.11903 · Cited by 4,263 · cited 1×
“Note that this type of method can only be used when prior knowledge is known in advance about shortcuts.”
MethodRight 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 2×
“NLI, QA, and reading comprehension (Zhou and Bansal, 2020; McCoy et al., 2019).”
Arjovsky, M., Bottou, L., Gulrajani, I., & Lopez-Paz, D. (2019) · arXiv · doi:10.48550/arxiv.1907.02893 · Cited by 927 · cited 3×
“This is distinct from the literature that works with the toy task (e.g., Colored MNIST [1]), which typically contains a single type of bias and the bias is fully known.”
BackgroundMethodInfluential
Research works printed but not linked: 14
- Branco, R., Branco, A., Silva, J., and Rodrigues, J. Shortcutted commonsense: Data spuriousness in deep learning of commonsense reasoning. In Proceedings of the 2021 Conf. Empirical Methods in Natural Language Processing.
- Devlin, J., Chang, M., Lee, K., and Toutanova, K. Bert: Pre-training of deep bidirectional transformers for language understanding. North American Chapter of the Assoc. Computational Linguistics, 2019.
- Ethayarajh, K. How contextual are contextualized word representations? Comparing the geometry of BERT, ELMo, and GPT-2 embeddings. In Proceedings of the 2019 Conf. Empirical Methods in Natural Language Processing and the 9th Intern. Joint Conf. Natural Language Processing, 55--65.
- Gururangan, S., Swayamdipta, S., Levy, O., Schwartz, R., Bowman, S.R., and Smith, N.A. Annotation artifacts in natural language inference data. North American Chapter of the Assoc. Computational Linguistics, 2018.
- Liu, Y. et al. RoBERTa: A robustly optimized BERT pretraining approach, 2019; arXiv:1907.11692. · arXiv
- Robinson, J., Sun, L., Yu, K., Batmanghelich, K., Jegelka, S., and Sra, S. Can contrastive learning avoid shortcut solutions? Advances in Neural Information Processing Systems, 2021.
- Sen, P. and Saffari, A. What do models learn from question answering datasets? In Proceedings of the 2020 Conf. Empirical Methods in Natural Language Processing.
- Si, C. et al. Prompting gpt-3 to be reliable, 2022; arXiv:2210.09150. · arXiv
- Stacey, J., Belinkov, Y., and Rei, M. Supervising Model Attention with Human Explanations for Robust Natural Language Inference. In Proceedings of the 2022 AAAI Conf. Artificial Intelligence.
- Sundararajan, M., Taly, A., and Yan, Q. Axiomatic attribution for deep networks. In Proceedings of the 2017 Intern. Conf. Machine Learning.
- Tu, L., Lalwani, G., Gella, S., and He, H. An empirical study on robustness to spurious correlations using pre-trained language models. Trans. Assoc. Computational Linguistics (2020).
- Webson, A. and Pavlick, E. Do prompt-based models really understand the meaning of their prompts? In Proceedings of the 2022 Conf. North American Chapter of the Assoc. Computational Linguistics: Human Language Technologies.
- Zellers, R., Bisk, Y., Schwartz, R., and Choi, Y. Swag: A large-scale adversarial dataset for grounded commonsense inference. In Proceedings of the 2018 Conf. Empirical Methods in Natural Language Processing.
- Zhao, Z., Wallace, E., Feng, S., Klein, D., and Singh, S. Calibrate before use: Improving few-shot performance of language models. In Proceedings of the 2021 Intern. Conf. Machine Learning, 12697--12706.
Cited by
The rise and potential of large language model based agents: a survey
Xi, Z., Chen, W., Guo, X., He, W., Ding, Y., Hong, B., Zhang, M., Wang, J., Jin, S., Zhou, E., et al. (2025) · Science China Information Sciences · doi:10.1007/s11432-024-4222-0 · Cited by 970
“Furthermore, employing debiasing methods and calibration techniques can also mitigate the potential fairness issues within LLMs [551,552].”
(2024) · doi:10.1109/access.2024.3365742 · Cited by 934
Explainability for Large Language Models: A Survey
Zhao, H., Chen, H., Yang, F., Liu, N., Deng, H., Cai, H., Wang, S., Yin, D., & Du, M. (2024) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3639372 · Cited by 681
“Post-hoc explainability methods can be used to analyze model feature importance patterns to identify biases or limitations in its behavior (Du et al., 2023).”
BackgroundMethodSelf-citationHarnessing the power of llms in practice: a survey on chatGPT and beyond
Yang, J., Jin, H., Tang, R., Han, X., Feng, Q., Jiang, H., Zhong, S., Yin, B., & Hu, X. (2024) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3649506 · Cited by 644
Self-citationA Survey on Large Language Models: Applications, Challenges, Limitations, and Practical Usage
Hadi, M. U., tashi, Q. A., Qureshi, R., Shah, A., muneer, A., Irfan, M., Zafar, A., Shaikh, M. B., Akhtar, N., Wu, J., et al. (2023) · doi:10.36227/techrxiv.23589741.v1 · Cited by 370
A survey on LLM-based multi-agent systems: Workflow, infrastructure, and challenges
Li, X., Wang, S., Zeng, S., Wu, Y., & Yang, Y. (2024) · Vicinagearth · doi:10.1007/s44336-024-00009-2 · Cited by 404
“Shortcut learning can be mitigated through methods such as data debias-ing, adversarial training, interpretive regularization, and confidence regularization [213].”
Foundational Challenges in Assuring Alignment and Safety of Large Language Models
Anwar, U., Saparov, A., Ramírez, J. R., Paleka, D., Turpin, M., Hase, P., Lubana, E. S., Jenner, E., Casper, S., Sourbut, O., et al. (2024) · arXiv (Cornell University) · doi:10.3929/ethz-b-000706962 · Cited by 17
Large language models in law: A survey
Lai, J., Gan, W., Wu, J., Qi, Z., & Yu, P. S. (2024) · AI Open · doi:10.1016/j.aiopen.2024.09.002 · Cited by 359
Hadi, M. U., tashi, Q. A., Qureshi, R., Shah, A., muneer, A., Irfan, M., Zafar, A., Shaikh, M. B., Akhtar, N., Wu, J., et al. (2023) · doi:10.36227/techrxiv.23589741.v4 · Cited by 103
Source Code Summarization in the Era of Large Language Models
Sun, W., Miao, Y., Li, Y., Zhang, H., Fang, C., Liu, Y., Deng, G., Liu, Y., & Chen, Z. (2025) · ACM/IEEE International Conference on Software Engineering: Companion Proceedings (ICSE-Companion) · doi:10.1109/icse55347.2025.00034 · Cited by 213
A comprehensive review of large language models: issues and solutions in learning environments
Shahzad, T., Mazhar, T., Tariq, M. U., Ahmad, W., Ouahada, K., & Hamam, H. (2025) · Discover Sustainability · doi:10.1007/s43621-025-00815-8 · Cited by 176
“These models may be identified by examining large amounts of text [55].”
Niu, Q., Liu, J., Bi, Z., Feng, P., Peng, B., Chen, K., Li, M., Yan, L. K. Q., Zhang, Y., Yin, C. H., et al. (2026) · BIO Integration · doi:10.15212/bioi-2025-0199 · Cited by 121
Hadi, M. U., tashi, Q. A., Qureshi, R., Shah, A., muneer, A., Irfan, M., Zafar, A., Shaikh, M. B., Akhtar, N., Wu, J., et al. (2023) · doi:10.36227/techrxiv.23589741 · Cited by 83
Language Model Behavior: A Comprehensive Survey
Chang, T. A., & Bergen, B. K. (2024) · Computational Linguistics · doi:10.1162/coli_a_00492 · Cited by 105
“During the fine-tuning process, language models are prone to overfitting to spurious correlations between text features and labels in the fine-tuning dataset (McCoy, Pavlick, and Linzen 2019; Kavumba et al. 2020; Wang et al. 2022b; Du et al. 2022a; Kavumba, Takahashi, and Oda 2022), and they can even “forget” syntacti…”
BackgroundGuizani, S., Mazhar, T., Shahzad, T., Ahmad, W., Bibi, A., & Hamam, H. (2025) · Discover Education · doi:10.1007/s44217-025-00424-7 · Cited by 114
From Vulnerability to Defense: The Role of Large Language Models in Enhancing Cybersecurity
Kasri, W., Himeur, Y., Alkhazaleh, H. A., Tarapiah, S., Atalla, S., Mansoor, W., & Al-Ahmad, H. (2025) · Computation · doi:10.3390/computation13020030 · Cited by 70
DISCO: Distilling Counterfactuals with Large Language Models
Chen, Z., Gao, Q., Bosselut, A., Sabharwal, A., & Richardson, K. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.302 · Cited by 160
“, 2020), dataset biases continue to be a formidable challenge for robust model development (Gururangan et al., 2018; Poliak et al., 2018; Kaushik and Lipton, 2018; Tsuchiya, 2018; Liu et al., 2020b; Du et al., 2022).”
BackgroundHadi, M. U., tashi, Q. A., Qureshi, R., Shah, A., muneer, A., Irfan, M., Zafar, A., Shaikh, M. B., Akhtar, N., Wu, J., et al. (2023) · doi:10.36227/techrxiv.23589741.v3 · Cited by 53
RecRanker: Instruction Tuning Large Language Model as Ranker for Top-k Recommendation
Luo, S., He, B., Zhao, H., Shao, W., Qi, Y., Huang, Y., Zhou, A., Yao, Y., Li, Z., Xiao, Y., et al. (2024) · ACM Transactions on Information Systems · doi:10.1145/3705728 · Cited by 108
“Consequently, they have been increasingly integrated into a multitude of domains, including recommender systems [18]–[20].”
Wu, X., Saraf, P. P., Lee, G., Latif, E., Liu, N., & Zhaı, X. (2025) · Technology Knowledge and Learning · doi:10.1007/s10758-025-09836-8 · Cited by 62
“These unexpected paths usually are referred to shortcuts [5, 17], indicating that LLMs find a superficial way to make predictions without following a reasonable logic chain as human experts.”
A Survey of Test-time Compute: From Intuitive Inference to Deliberate Reasoning
Ji, Y., Li, J., Xiang, Y., Ye, H., Wu, K., Kai, Y., Xu, J., Mo, L., & Zhang, M. (2026) · Computational Linguistics · doi:10.1162/coli.a.629 · Cited by 64
OmniSQL: Synthesizing High-Quality Text-to-SQL Data at Scale
Li, H., Wu, S., Zhang, X., Huang, X., Zhang, J., Jiang, F., Wang, S., Zhang, T., Chen, J., Shi, R., et al. (2025) · Proceedings of the VLDB Endowment · doi:10.14778/3749646.3749723 · Cited by 56
Testing AI on language comprehension tasks reveals insensitivity to underlying meaning
Dentella, V., Günther, F., Murphy, E., Marcus, G., & Leivada, E. (2024) · Scientific Reports · doi:10.1038/s41598-024-79531-8 · Cited by 54
“While this issue was ignored for the purposes of coding in the one-word setting, because we wanted to give the models the benefit of the doubt, it further showcases how LLMs’ outputs are sensitive to the meaning of prompts only in a superficial manner 61 .”
Zheng, Y.-B., Zhou, Y.-X., Chen, X.-D., & Ye, X.-D. (2025) · Computer Assisted Language Learning · doi:10.1080/09588221.2025.2453191 · Cited by 41
Hadi, M. U., Al-Tashi, Q., Qureshi, R., Shah, A., Muneer, A., Irfan, M., Zafar, A., Shaikh, M. B., Akhtar, N., Hassan, S. Z., et al. (2023) · doi:10.36227/techrxiv.23589741.v2 · Cited by 142
Can Language Models Learn to Skip Steps?
Liu, T., Guo, Q., Hu, X., Jiayang, C., Zhang, Y., Qiu, X., & Zhang, Z. (2024) · neural information processing systems · doi:10.52202/079017-1441 · Cited by 106
Self-citationHadi, M. U., Tashi, Q. A., Qureshi, R., Shah, A., Muneer, A., Irfan, M. I., Zafar, A., Shaikh, M. B., Akhtar, N., Hassan, S. Z., et al. (2025) · doi:10.36227/techrxiv.23589741.v8 · Cited by 59
Quttainah, M. A., Mishra, V., Madakam, S., Lurie, Y., & Mark, S. (2024) · JMIR AI · doi:10.2196/51834 · Cited by 61
“…harm to user reputation Fairness E4 [27,38] Taking responsibility for the obligation to treat users with honesty and morality Accountability E5 [27,30] Functioning in a way that makes it simple for others to observe what actions are taken Transparency E6 [29,30] Ability to describe how the models…”
BackgroundPternea, M., Singh, P., Chakraborty, A., Oruganti, Y. D., Milletari, M., Bapat, S., & Jiang, K. (2024) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.15960 · Cited by 59
“These studies use RL to improve the performance of the LLM in an NLP task.”
MethodFostering effective hybrid human-LLM reasoning and decision making
Passerini, A., Gema, A. P., Minervini, P., Sayin, B., & Tentori, K. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2024.1464690 · Cited by 51
“One consequence is that, like many other sub-symbolic machine learning models, LLMs are prone to shortcut learning (Du et al., 2023), a tendency to rely on non-robust features that are spuriously correlated with ground-truth supervision in the training data, yet fail to generalize in…”
Shortcut Learning Explanations for Deep Natural Language Processing: A Survey on Dataset Biases
Dogra, V., Verma, S., Kavita, K., Woźniak, M., Shafi, J., & Ijaz, M. F. (2024) · IEEE Access · doi:10.1109/access.2024.3360306 · Cited by 46
“From the examples, it is clear that the performance and behavior of NLP models are significantly influenced by the datasets that are used, as both researchers and practitioners have come to understand over time [15].”
BackgroundCMAT: A Multi-Agent Collaboration Tuning Framework for Enhancing Small Language ModelsPreprint
Liang, X., He, Y., Tao, M., Xia, Y., Wang, J., Shi, T., Wang, J., & Yang, J. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2404.01663 · Cited by 43
Distributed Training of Large Language Models
Zeng, F., Gan, W., Wang, Y., & Yu, P. S. (2023) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads60453.2023.00126 · Cited by 55
Show Me the Work: Fact-Checkers' Requirements for Explainable Automated Fact-Checking
Warren, G., Shklovski, I., & Augenstein, I. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713277 · Cited by 66
Large Language Models Can be Lazy Learners: Analyze Shortcuts in In-Context Learning
Tang, R., Kong, D., Huang, L., & Xue, H. (2023) · Findings of the Association for Computational Linguistics: ACL 2023 · doi:10.18653/v1/2023.findings-acl.284 · Cited by 31
“non-robust knowledge and robust knowledge (Ilyas et al., 2019; Du et al., 2022).”
BackgroundRobustness Challenges in Model Distillation and Pruning for Natural Language Understanding
Du, M., Mukherjee, S., Cheng, Y., Shokouhi, M., Xia, H., & Awadallah, A. H. (2023) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.129 · Cited by 47
“Due to the independent and identically distributed (IID) split of training, development, and test sets, these models that learn spurious decision rules from training data can perform well on in-distribution data (Du et al., 2022).”
BackgroundSelf-citationEvaluating the performance and robustness of LLMs in materials science Q&A and property predictions
Wang, H., Li, K., Ramsay, S., Fehlis, Y., Kim, E., & Hattrick-Simpers, J. R. (2025) · Digital Discovery · doi:10.1039/d5dd00090d · Cited by 38
“This aligns with the shortcut learning behavior observed in LLMs, where models rely on superficial correlations rather than learning meaningful patterns from the data [48].”
Esale: Enhancing Code-Summary Alignment Learning for Source Code Summarization
Fang, C., Sun, W., Chen, Y., Chen, X., Wei, Z., Zhang, Q., You, Y., Luo, B., Liu, Y., & Chen, Z. (2024) · IEEE Transactions on Software Engineering · doi:10.1109/tse.2024.3422274 · Cited by 39
“Recently, with the success of large language models (LLMs) in NLP [93], [94], an increasing number of SE researchers have started integrating them into the resolution process of various SE tasks [95], [96], including code summarization tasks.”
BackgroundRectify ViT Shortcut Learning by Visual Saliency
Ma, C., Zhao, L., Chen, Y.-Z., Guo, L., Zhang, T., Hu, X.-T., Shen, D., Jiang, X., & Liu, T. (2022) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3310531 · Cited by 30
“In addition, among different strategies aiming to address shortcut learning, domain-specific knowledge is considered an important approach to avoiding the unintended representations/features learned by shortcut learning [17], [22], [57].”
Self-citationGuess the Instruction! Flipped Learning Makes Language Models Stronger Zero-Shot LearnersPreprint
Ye, S., Kim, D., Jang, J., Shin, J., & Seo, M. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2210.02969 · Cited by 39
“We conjecture that this is a result of shortcut learning of large LMs (Du et al., 2022; Min et al., 2022c).”
BackgroundCan language representation models think in bets?
Tang, Z., & Kejriwal, M. (2023) · Royal Society Open Science · doi:10.1098/rsos.221585 · Cited by 23
“It is important to design the prompts to control for a range of problems that are known to occur with such language models, including their sensitivity to format [78], and their propensity to achieve high performance through advanced, but ultimately superficial, statistical pattern matching [42].”
BackgroundSLANG: New Concept Comprehension of Large Language Models
Mei, L., Liu, S., Wang, Y., Bi, B., & Cheng, X. (2024) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2024.emnlp-main.698 · Cited by 56
Shortcut Learning in In-Context Learning: A SurveyPreprint
Song, R., Li, Y., Shi, L., Giunchiglia, F., & Xu, H. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2411.02018 · Cited by 26
Robustness in Large Language Models: A Survey of Mitigation Strategies and Evaluation Metrics
Kumar, P., & Mishra, S. (2025) · Transactions on machine learning research · doi:10.48550/arxiv.2505.18658 · Cited by 29
“Prompt Robustness Frequency of the most common claim across (RC) (Rahman et al., 2024) paraphrased prompts.”
Li, J.-H., Xu, J.-M., Li, S.-Y., Huang, S., Liu, J., Lian, Y., & Dai, G.-H. (2023) · International Conference on Computer Aided Design · doi:10.1145/3676536.3676796 · Cited by 19
Mechanistic indicators of understanding in large language models
Beckmann, P., & Queloz, M. (2026) · Philosophical Studies · doi:10.1007/s11098-026-02513-1 · Cited by 31
“…that LLMs can memorize training fragments verbatim (Carlini et al., 2023; Aerni et al., 2024; Stoehr et al., 2024), detect n -grams and learn n -gram statistics (Svete et al., 2024; Voita et al., 2024; Chang, 2025), and discover heuristics exploiting shallow correlations (see Du et al., 2023).”
Efficient Compression of Large Language Models: A Case Study on Llama 2 with 13B ParametersPreprint
Konishi, M., Nakano, K., & Tomoda, Y. (2024) · Research Square · doi:10.21203/rs.3.rs-4407954/v1 · Cited by 64
From Form(s) to Meaning: Probing the Semantic Depths of Language Models Using Multisense Consistency
Ohmer, X. I., Bruni, E., & Hupkes, D. (2024) · Computational Linguistics · doi:10.1162/coli_a_00529 · Cited by 36
“Adversarial datasets (e.g., Nie et al. 2020) are designed precisely to expose such shortcut learning behaviors (for an overview of shortcut learning, see Du et al. 2023).”
Can ChatGPT Perform Reasoning Using the IRAC Method in Analyzing Legal Scenarios Like a Lawyer?
Kang, X., Qu, L., Soon, L.-K., Trakic, A., Zhuo, T. Y., Emerton, P., & Grant, G. M. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.929 · Cited by 33
“LLMs are known to rely on “shortcuts” for predictions occasionally (Du et al., 2022), hence their robustness is questionable.”
BackgroundNiu, Q., Liu, J., Bi, Z., Feng, P., Peng, B., Chen, K., Li, M., Yan, L. K., Zhang, Y., Yin, C. H., et al. (2024) · arXiv · doi:10.48550/arxiv.2409.02387 · Cited by 16
“Du et al. [57] reviewed recent developments addressing shortcut learning and robustness challenges in LLMs, suggesting the combination of data-driven schemes with domain knowledge and the introduction of more inductive biases into model architectures.”
Moëll, B., Farestam, F., & Beskow, J. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2025.1557920 · Cited by 20
A Gradient Control Method for Backdoor Attacks on Parameter-Efficient Tuning
Gu, N., Fu, P., Liu, X., Liu, Z., Lin, Z., & Wang, W. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.194 · Cited by 37
“It can be seen that the difficulty of model training for the two kinds of data is different, and the trigger in the poisoned data is a recurring feature, which is easier for the model to recognize (Du et al., 2022a).”
BackgroundKaleidoscope: In-language Exams for Massively Multilingual Vision EvaluationPreprint
Salazar, I., Burda, M. F., Islam, S. B., Moakhar, A. S., Singh, S., Farestam, F., Romanou, A., Boiko, D., Khullar, D., Zhang, M., et al. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.07072 · Cited by 17
Shukor, M., Ramé, A., Dancette, C., & Cord, M. (2023) · International Conference on Learning Representations · doi:10.48550/arxiv.2310.00647 · Cited by 18
“For example, performances on those tasks may artificially increase simply by exploiting dataset biases and shortcuts, without truly understanding and generalization (Geirhos et al., 2020; Dancette et al., 2021; Du et al., 2022).”
BackgroundLaurer, M., van Atteveldt, W., Casas, A., & Welbers, K. (2024) · Communication Methods and Measures · doi:10.1080/19312458.2024.2378690 · Cited by 13
“There is a broad literature on how machine learning models rely on simple patterns (shortcuts) in their (pre)training data to solve their tasks, instead of truly understanding the underlying task and concept (Du et al., 2022).”
An Overview on Large Language Models
Dakhel, A. M., Nikanjam, A., Khomh, F., Desmarais, M. C., & Washizaki, H. (2024) · Springer eBooks · doi:10.1007/978-3-031-55642-5_1 · Cited by 15
FactAlign: Fact-Level Hallucination Detection and Classification Through Knowledge Graph Alignment
Rashad, M., Zahran, A., Amin, A., Abdelaal, A., & Altantawy, M. (2024) · Workshop on Trustworthy Natural Language Processing (TrustNLP) · doi:10.18653/v1/2024.trustnlp-1.8 · Cited by 23
A Survey on Measuring and Mitigating Reasoning Shortcuts in Machine Reading ComprehensionPreprint
Ho, X., Meissner, J. M., Sugawara, S., & Aizawa, A. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2209.01824 · Cited by 12
“Although there are some survey papers on reasoning shortcuts (Geirhos et al., 2020; Schlegel et al., 2020a; Wang et al., 2022b; Du et al., 2023), they predominantly focus on general tasks in NLP or in larger domains such as computer vision.”
BackgroundInfluentialExploring Robustness of Multilingual LLMs on Real-World Noisy DataPreprint
Aliakbarzadeh, A., Flek, L., & Karimi, A. (2025) · Qeios · doi:10.32388/3x6cxv · Cited by 14
COMI: COrrect and MItigate Shortcut Learning Behavior in Deep Neural Networks
Zhao, L., Liu, Q., Yue, L., Chen, W., Chen, L., Sun, R., & Song, C. (2024) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3626772.3657729 · Cited by 23
Self-Supervised Position Debiasing for Large Language Models
Liu, Z., Chen, Z., Zhang, M., Ren, Z., Ren, P., & Chen, Z. (2024) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2024.findings-acl.170 · Cited by 11
Examination of Code generated by Large Language ModelsPreprint
Beer, R., Feix, A., Guttzeit, T., Muras, T., Müller, V. C., Rauscher, M., Schäffler, F., & Löwe, W. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2408.16601 · Cited by 10
“A Large Language Model is a type of artificial intelligence used for processing language [5].”
MethodUnderstanding Hallucinations in Large Visual and Language Models
Ho, Z. Y., Liang, S., & Tao, D. (2026) · ACM Computing Surveys · doi:10.1145/3811409 · Cited by 11
Self-citationLearning Shortcuts: On the Misleading Promise of NLU in Language Models
Bihani, G., & Rayz, J. T. (2025) · Springer eBooks · doi:10.1007/978-3-031-67886-8_9 · Cited by 8
Ethical perspectives on deployment of large language model agents in biomedicine: a survey
Nia, N. G., Amiri, A., Luo, Y., & Kline, A. S. (2025) · AI and Ethics · doi:10.1007/s43681-025-00847-w · Cited by 9
“Bias in LLMs can arise at multiple stages: data collection, pretraining objective design, model architecture, and finetuning procedures [ 103].”
CallNavi, A challenge and empirical study on LLM function calling and routing
Song, Y., Tang, X., Lothritz, C., Ezzini, S., Klein, J., Bissyandé, T. F., Boytsov, A., Ble, U., & Goujon, A. (2025) · International Conference on Evaluation and Assessment in Software Engineering (EASE) · doi:10.1145/3756681.3756975 · Cited by 7
Natural Language Inference with Transformer Ensembles and Explainability Techniques
Perikos, I., & Souli, S. (2024) · Electronics · doi:10.3390/electronics13193876 · Cited by 6
“These models, such as BERT, RoBERTa, and DeBERTa, have demonstrated significant improvements in NLI tasks by capturing intricate patterns in language through deep contextual embeddings [8].”
Huang, S., Li, J.-H., Yu, Z., Ye, J.-C., Xu, J.-M., Xu, N.-Y., & Dai, G.-H. (2025) · Asia and South Pacific Design Automation Conference · doi:10.1145/3658617.3697618 · Cited by 11
Adaptive Evolutionary Computing Ensemble Learning Model for Sentiment Analysis
Liu, X., Zhang, K., Fiumara, G., Meo, P. D., & Ficara, A. (2024) · Applied Sciences · doi:10.3390/app14156802 · Cited by 8
Understanding Large Language Models
Sreerakuvandana, S., Pappachan, P., & Arya, V. (2024) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3693-3860-5.ch001 · Cited by 7
Automated Trustworthiness Oracle Generation for Machine Learning Text Classifiers
Tung, L. N., Cho, S. B., Du, X., Neelofar, N., Terragni, V., Ruberto, S., & Aleti, A. (2025) · Proceedings of the ACM on software engineering. · doi:10.1145/3729376 · Cited by 8
Emerging Threat Vectors: How Malicious Actors Exploit LLMs to Undermine Border Security
Doumanas, D., Karakikes, A., Soularidis, A., Mainas, E., & Kotis, K. (2025) · AI · doi:10.3390/ai6090232 · Cited by 5
“With prompt engineering, carefully crafted inputs can direct models to perform specific tasks without retraining [12–14].”
Arous, I., Chehbouni, K., Cheng, Z., & Dossou, B. F. P. (2025) · Springer eBooks · doi:10.1007/978-981-97-8440-0_85-1 · Cited by 6
Learning from shortcut: a shortcut-guided approach for explainable graph learning
Yue, L., Liu, Q., Liu, Y., Gao, W., & Yao, F. (2025) · Frontiers of Computer Science · doi:10.1007/s11704-024-40452-4 · Cited by 5
Large Language Model on Multi-Modal Data
Aneja, A., Dhull, A., Singh, A., & Singh, K. K. (2025) · Springer eBooks · doi:10.1007/978-981-96-2355-6_3 · Cited by 8
Sheik, R., Reji, S. A., Sharon, A., Avisha, M., & Nirmala, S. J. (2025) · Artificial Intelligence and Law · doi:10.1007/s10506-025-09459-5 · Cited by 5
Gradient Extrapolation for Debiased Representation Learning
Asaad, I., Shadaydeh, M., & Denzler, J. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00364 · Cited by 5
“These correlations often cause models trained with Empirical Risk Minimization (ERM) [45] to rely on these correlations instead of the true, intrinsic features of the classes [10, 12, 40].”
ELF-Gym: Evaluating Large Language Models Generated Features for Tabular Prediction
Zhang, Y., Li, N., Gan, Q., Zhang, W., Wipf, D., & Wang, M. (2024) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3627673.3679153 · Cited by 4
Exploring and Mitigating Shortcut Learning for Generative Large Language Models
Sun, Z., Xiao, Y., Li, J., Ji, Y., Chen, W., & Zhang, M. (2024) · Language Resources and Evaluation Conference (LREC) · doi:10.63317/4z4vjvv2kxaj · Cited by 28
Boosting LLM-based Relevance Modeling with Distribution-Aware Robust Learning
Liu, H., Gong, S., Ji, Y., Wu, K., Xu, J., & Gu, J. (2024) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3627673.3680052 · Cited by 8
SPrint: Self-Paced Continual Learning with Adaptive Curriculum and Memory Replay
Kim, M.-S., Liu, L., & Kwon, H.-Y. (2024) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata62323.2024.10825931 · Cited by 4
Controlling Learned Effects to Reduce Spurious Correlations in Text Classifiers
Bansal, P., & Sharma, A. K. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.127 · Cited by 8
Does Deep Learning Learn to Abstract? A Systematic Probing FrameworkPreprint
An, S., Lin, Z., Chen, B., Fu, Q., Zheng, N., & Lou, J. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2302.11978 · Cited by 6
“…of deep learning may come from the memorization of some surface patterns (also called superficial correlations or shortcuts) (Geirhos et al., 2020; Du et al., 2022), such as some special tokens (Niven & Kao, 2020; Gururangan et al., 2018), overlapping contexts (Lai et al., 2021; Sen & Saffari,…”
BackgroundConditional Language Learning with ContextPreprint
Zhang, X., Li, M., & Wu, J. Y. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2406.01976 · Cited by 4
“Moreover, if the property c is simple, the model may also adapt p ( c ) faster and earlier than learning p ( x c ) (Geirhos et al., 2020; Du et al., 2024).”
BackgroundBae, S., Choi, Y., & Lee, J. (2025) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2025.naacl-long.624 · Cited by 6
Su, S., Tanaka, N., Ushiku, Y., & Takahashi, K. (2026) · ACS Applied Materials & Interfaces · doi:10.1021/acsami.5c25223 · Cited by 3
Advances in reasoning by prompting large language models: A survey
Hong, R., Pang, X., & Zhang, C. (2026) · Cybernetics and Intelligence · doi:10.26599/cai.2024.9390004 · Cited by 11
Language-Based Augmentation to Address Shortcut Learning in Object-Goal Navigation
Hoftijzer, D., Burghouts, G., & Spreeuwers, L. (2023) · 2023 Seventh IEEE International Conference on Robotic Computing (IRC) · doi:10.1109/irc59093.2023.00007 · Cited by 2
“Recent works show that even Large-Language Models (LLMs), pre-trained on text amounting to billions of words, suffer from shortcut learning, largely due to collection artifacts in training data [7], [21].”
Background(2023) · Journal of System and Management Sciences · doi:10.33168/jsms.2023.0433 · Cited by 2
“Consequently, the RoBERTa algorithm emerged as a superior alternative, boasting enhanced performance and reduced computational demands (Dillan & Fudholi, 2022; Du et al., 2022; Liao et al., 2021).”
InfluentialAutomatic Assessment of Active Learning in Online Discussions with Large Language Models
Techawitthayachinda, R., & Iriya, R. (2024) · Lecture notes on data engineering and communications technologies · doi:10.1007/978-981-97-9255-9_3 · Cited by 2
MiMu: mitigating multiple shortcut learning behavior of transformers
Zhao, L., Liu, Q., Chen, W., Chen, L., Sun, R., Hou, M., Wang, Y., Wang, S., Ren, P., & Zhou, J. (2025) · Frontiers of Computer Science · doi:10.1007/s11704-025-50448-3 · Cited by 2
Mitigating Shortcuts in Language Models with Soft Label Encoding
He, Z., Deng, H., Zhao, H., Liu, N., & Du, M. (2024) · Language Resources and Evaluation Conference (LREC) · doi:10.63317/58mojxfrnf6g · Cited by 2
Self-citationHuang, J., Yi, B., Sun, W., Wan, B., Xu, Y., Feng, Y., Ye, W., & Qin, Q. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.5002351 · Cited by 3
Do LLMs Overcome Shortcut Learning?An Evaluation of Shortcut Challenges in Large Language ModelsPreprint
Yuan, Y., Zhao, L., Zhang, K., Zheng, G., & Liu, Q. (2024) · Qeios · doi:10.32388/1bzh25 · Cited by 2
GIM: Evaluating models via tasks that integrate multiple cognitive domainsPreprint
Patel, R., Rezende, A., & McClain, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.18663 · Cited by 2
When and Why Does Bias Mitigation Work?
Ravichander, A., Stacey, J., & Rei, M. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.619 · Cited by 11
Continual debiasing: A bias mitigation framework for natural language understanding systems
Lee, M., Kim, J., Park, J.-H., & Lee, S. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.126593 · Cited by 1
The Reliability Paradox: Exploring How Shortcut Learning Undermines Language Model Calibration
Bihani, G., & Rayz, J. T. (2025) · Hawaii International Conference on System Sciences (HICSS) · doi:10.24251/hicss.2025.102 · Cited by 1
Sun, Y., & Huo, J. (2025) · Journal of Computational Methods in Sciences and Engineering · doi:10.1177/14727978251324129 · Cited by 1
“30,31 Acs J used multilingual datasets to detect morphological information in language models and found that the pre-trained BERT model performed strongly in these tasks.”
Language-Based Augmentation to Mitigate Shortcut Learning in Object-Goal Navigation
Hoftijzer, D., Burghouts, G. J., & Spreeuwers, L. (2025) · International Journal of Semantic Computing · doi:10.1142/s1793351x25410077 · Cited by 1
Metadata
- Authors
- Du, M., He, F., Zou, N., Tao, D., & Hu, X.
- Affiliations
- New Jersey Institute of Technology (US)
- University of Edinburgh (GB)
- Texas A&M University (US)
- The University of Sydney (AU)
- Rice University (US)
- Venue
- Communications of the ACM
- Published
- 2023-12-21
- Volume
- 67
- Issue
- 1
- Pages
- 110-120
- ISSN
- 0001-0782 (ISSN-L)
- 1557-7317
- OpenAlex
- W4390043316
- Cited by
- 94
- References
- 40
- Citations by year
- 2026: 22 · 2025: 34 · 2024: 22 · 2023: 16
- FWCI
- 8.86
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Communications of the ACM (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- 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
- shortcut learning, large language models, natural language understanding, robustness, model interpretability
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.