Can We Trust AI Benchmarks? An Interdisciplinary Review of Current Issues in AI Evaluation
The DOI matches its Crossref record; 19 of the 19 references it cites were found in a registry or an index.
Abstract
Quantitative Artificial Intelligence (AI) Benchmarks have emerged as fundamental tools for evaluating the performance, capability, and safety of AI models and systems. Currently, they shape the direction of AI development and are playing an increasingly prominent role in regulatory frameworks. As their influence grows, however, so too does concerns about how and with what effects they evaluate highly sensitive topics such as capabilities, including high-impact capabilities, safety and systemic risks. This paper presents an interdisciplinary meta-review of about 110 studies that discuss shortcomings in quantitative benchmarking practices, published in the last 10 years. It brings together many fine-grained issues in the design and application of benchmarks (such as biases in dataset creation, inadequate documentation, data contamination, and failures to distinguish signal from noise) with broader sociotechnical issues (such as an over-focus on evaluating text-based AI models according to one-time testing logic that fails to account for how AI models are increasingly multimodal and interact with humans and other technical systems). Our review also highlights a series of systemic flaws in current benchmarking practices, such as misaligned incentives, construct validity issues, unknown unknowns, and problems with the gaming of benchmark results. Furthermore, it underscores how benchmark practices are fundamentally shaped by cultural, commercial and competitive dynamics that often prioritise state-of-the-art performance at the expense of broader societal concerns. By providing an overview of risks associated with existing benchmarking procedures, we problematise disproportionate trust placed in benchmarks and contribute to ongoing efforts to improve the accountability and relevance of quantitative AI benchmarks within the complexities of real-world scenarios.
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
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.
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
Fairness and Abstraction in Sociotechnical Systems
Selbst, A. D., Boyd, D., Friedler, S. A., Venkatasubramanian, S., & Vertesi, J. (2019) · Proceedings of the Conference on Fairness, Accountability, and Transparency · doi:10.1145/3287560.3287598 · Cited by 1,631
Systematic literature studies: database searches vs. backward snowballing
Jalali, S., & Wohlin, C. (2012) · Proceedings of the ACM-IEEE international symposium on Empirical software engineering and measurement · doi:10.1145/2372251.2372257 · Cited by 659
Benchmarking in Optimization: Best Practice and Open Issues
Bartz-Beielstein, T., Doerr, C., Berg, D. V. D., Bossek, J., Chandrasekaran, S., Eftimov, T., Fischbach, A., Kerschke, P., La Cava, W., Lopez-Ibanez, M., et al. (2020) · arXiv · doi:10.48550/arxiv.2007.03488 · Cited by 214
The growing influence of industry in AI research
Ahmed, N., Wahed, M., & Thompson, N. C. (2023) · Science · doi:10.1126/science.ade2420 · Cited by 208
AI auditing: The Broken Bus on the Road to AI Accountability
Birhane, A., Steed, R., Ojewale, V., Vecchione, B., & Raji, I. D. (2024) · 2024 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) · doi:10.1109/satml59370.2024.00037 · Cited by 158
Utility is in the Eye of the User: A Critique of NLP Leaderboards
Ethayarajh, K., & Jurafsky, D. (2020) · Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2020.emnlp-main.393 · Cited by 147
Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Text
Gehrmann, S., Clark, E., & Sellam, T. (2023) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.13715 · Cited by 123
Rethink reporting of evaluation results in AI
Burnell, R., Schellaert, W., Burden, J., Ullman, T. D., Martinez-Plumed, F., Tenenbaum, J. B., Rutar, D., Cheke, L. G., Sohl-Dickstein, J., Mitchell, M., et al. (2023) · Science · doi:10.1126/science.adf6369 · Cited by 104
Inadequacies of Large Language Model Benchmarks in the Era of Generative Artificial Intelligence
McIntosh, T. R., Susnjak, T., Arachchilage, N., Liu, T., Watters, P., & Halgamuge, M. N. (2024) · arXiv · doi:10.48550/arxiv.2402.09880 · Cited by 104
Bringing the People Back In: Contesting Benchmark Machine Learning Datasets
Denton, R., Hanna, A., Amironesei, R., Smart, A., Nicole, H., & Scheuerman, M. K. (2020) · arXiv · doi:10.48550/arxiv.2007.07399 · Cited by 100
Scalable Extraction of Training Data from (Production) Language Models
Nasr, M., Carlini, N., Hayase, J., Jagielski, M., Cooper, A. F., Ippolito, D., Choquette-Choo, C. A., Wallace, E., Tramèr, F., & Lee, K. (2023) · arXiv · doi:10.48550/arxiv.2311.17035 · Cited by 96
Mapping global dynamics of benchmark creation and saturation in artificial intelligence
Ott, S., Barbosa-Silva, A., Blagec, K., Brauner, J., & Samwald, M. (2022) · Nature Communications · doi:10.1038/s41467-022-34591-0 · Cited by 73
Dehghani, M., Tay, Y., Gritsenko, A. A., Zhao, Z., Houlsby, N., Diaz, F., Metzler, D., & Vinyals, O. (2021) · arXiv · doi:10.48550/arxiv.2107.07002 · Cited by 48
Ground truth tracings (GTT): On the epistemic limits of machine learning
Kang, E. B. (2023) · Big Data & Society · doi:10.1177/20539517221146122 · Cited by 47
Evaluating the World Model Implicit in a Generative Model
Vafa, K., Chen, J., Rambachan, A., Kleinberg, J., & Mullainathan, S. (2024) · Advances in Neural Information Processing Systems 37 · doi:10.52202/079017-0846 · Cited by 22
Questionable practices in machine learning
Leech, G., Vazquez, J. J., Kupper, N., Yagudin, M., & Aitchison, L. (2024) · arXiv · doi:10.48550/arxiv.2407.12220 · Cited by 21
AI Sandbagging: Language Models can Strategically Underperform on Evaluations
van der Weij, T., Hofstätter, F., Jaffe, O., Brown, S. F., & Ward, F. R. (2024) · arXiv · doi:10.48550/arxiv.2406.07358 · Cited by 5
Data Contamination Through the Lens of Time
Roberts, M., Thakur, H., Herlihy, C., White, C., & Dooley, S. (2023) · arXiv · doi:10.48550/arxiv.2310.10628 · Cited by 2
Gema, A. P., Leang, J. O. J., Hong, G., Devoto, A., Mancino, A. C. M., Saxena, R., He, X., Zhao, Y., Du, X., Madani, M. R. G., et al. (2024) · arXiv · doi:10.48550/arxiv.2406.04127 · Cited by 1
Cited by
Developing the HAKI model as a conceptual framework for AI-assisted academic research
Shahini, A. (2026) · Discover Education · doi:10.1007/s44217-026-02041-4
Metadata
- Authors
- Eriksson, M., Purificato, E., Noroozian, A., Vinagre, J., Chaslot, G., Gomez, E., & Fernandez-Llorca, D.
- Venue
- Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society
- Published
- 2025-10-15
- Volume
- 8
- Issue
- 1
- Pages
- 850-864
- ISSN
- 3065-8365
- Cited by
- 26
- References
- 19
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.