Characteristics of Harmful Text: Towards Rigorous Benchmarking of Language Models
Open accessThe DOI matches its OpenAlex record; 20 of the 20 references it cites were found in a registry or an index.
Abstract
Large language models produce human-like text that drive a growing number of applications. However, recent literature and, increasingly, real world observations, have demonstrated that these models can generate language that is toxic, biased, untruthful or otherwise harmful. Though work to evaluate language model harms is under way, translating foresight about which harms may arise into rigorous benchmarks is not straightforward. To facilitate this translation, we outline six ways of characterizing harmful text which merit explicit consideration when designing new benchmarks. We then use these characteristics as a lens to identify trends and gaps in existing benchmarks. Finally, we apply them in a case study of the Perspective API, a toxicity classifier that is widely used in harm benchmarks. Our characteristics provide one piece of the bridge that translates between foresight and effective evaluation.
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
A Survey on Bias and Fairness in Machine Learning
Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2019) · arXiv · doi:10.48550/arxiv.1908.09635 · Cited by 5,812
A Survey on Bias and Fairness in Machine Learning
Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2022) · ACM Computing Surveys · doi:10.1145/3457607 · Cited by 5,812
Transformer-XL: Attentive language models beyond a fixed-length context
Dai, Z., Yang, Z., Yang, Y., Carbonell, J., Le, Q., & Salakhutdinov, R. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1285 · Cited by 3,169
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
Decolonial AI: Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence
Mohamed, S., Png, M.-T., & Isaac, W. (2020) · Philosophy & Technology · doi:10.1007/s13347-020-00405-8 · Cited by 912
Self-citationTruthfulQA: Measuring How Models Mimic Human Falsehoods
Lin, S., Hilton, J., & Evans, O. (2021) · arXiv · doi:10.48550/arxiv.2109.07958 · Cited by 899
RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models
Gehman, S., Gururangan, S., Sap, M., Choi, Y., & Smith, N. A. (2020) · Findings of the Association for Computational Linguistics: EMNLP 2020 · doi:10.18653/v1/2020.findings-emnlp.301 · Cited by 693
Gender bias in coreference resolution: Evaluation and debiasing methods
Zhao, J., Wang, T., Yatskar, M., Ordonez, V., & Chang, K.-W. (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/n18-2003 · Cited by 684
StereoSet: Measuring stereotypical bias in pretrained language models
Nadeem, M., Bethke, A., & Reddy, S. (2020) · arXiv · doi:10.48550/arxiv.2004.09456 · Cited by 643
Ex Machina: Personal Attacks Seen at Scale
Wulczyn, E., Thain, N., & Dixon, L. (2017) · Proceedings of the 26th International Conference on World Wide Web · doi:10.1145/3038912.3052591 · Cited by 636
Cited by
Auditing large language models: a three-layered approach
Mökander, J., Schuett, J., Kirk, H. R., & Floridi, L. (2024) · AI and Ethics · doi:10.1007/s43681-023-00289-2 · Cited by 289
When Can LLMs Actually Correct Their Own Mistakes? A Critical Survey of Self-Correction of LLMs
Kamoi, R., Zhang, Y., Zhang, N., Han, J., & Zhang, R. (2024) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00713 · Cited by 200
Generative Echo Chamber? Effect of LLM-Powered Search Systems on Diverse Information Seeking
Sharma, N., Liao, Q. V., & Xiao, Z. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642459 · Cited by 132
STELA: a community-centred approach to norm elicitation for AI alignment
Bergman, S., Marchal, N., Mellor, J., Mohamed, S., Gabriel, I., & Isaac, W. (2024) · Scientific Reports · doi:10.1038/s41598-024-56648-4 · Cited by 48
Self-citationLanguage Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey
Kumar, S., Balachandran, V., Njoo, L., Anastasopoulos, A., & Tsvetkov, Y. (2023) · Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.241 · Cited by 38
Metadata
- Authors
- Rauh, M., Mellor, J., Uesato, J., Huang, P.-S., Welbl, J., Weidinger, L., Dathathri, S., Glaese, A., Irving, G., Gabriel, I., Isaac, W. S., & Hendricks, L. A.
- Venue
- neural information processing systems
- Published
- 2022-01-01
- Pages
- 24720-24739
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W7133192930
- Cited by
- 14
- References
- 20
- FWCI
- 1.65
- Type
- conference paper
- Language
- en
- Open access
- green
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Characteristics of Harmful Text: Towards Rigorous Benchmarking of Language Models. https://app.cochat.ai/papers/10.52202/068431-1793
- The paper
- Rauh, M., Mellor, J., Uesato, J., Huang, P.-S., Welbl, J., Weidinger, L., Dathathri, S., Glaese, A., Irving, G., Gabriel, I., Isaac, W. S., & Hendricks, L. A. (2022). Characteristics of Harmful Text: Towards Rigorous Benchmarking of Language Models. neural information processing systems. https://doi.org/10.52202/068431-1793
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.