---
title: "SMARTER — paper by Nghiem et al. (2026) — summary and ideas"
description: "SMARTER is a two-stage framework for toxicity detection in low-data settings. 67 references, cited by 1 paper here."
authors: ["Huy Nghiem", "Advik Sachdeva", "Hal Daumé Iii"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.acl-long.1584/"
doi: "10.18653/v1/2026.acl-long.1584"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/nghiem-2026-smarter-data-efficient-framework-improve/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/nghiem-2026-smarter-data-efficient-framework-improve/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/17/) / SMARTER: A Data-efficient Framework to Improve Toxicity Detection with Explanation via Self-augmenting Large Language Models

# SMARTER: A Data-efficient Framework to Improve Toxicity Detection with Explanation via Self-augmenting Large Language Models

**Authors:** [Huy Nghiem](https://md.research.elamur.ai/free/en/authors/huy-nghiem/), Advik Sachdeva and [Hal Daumé Iii](https://md.research.elamur.ai/free/en/authors/hal-daume-iii/) · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.1584/) · [PDF](https://aclanthology.org/2026.acl-long.1584.pdf) · [DOI 10.18653/v1/2026.acl-long.1584](https://doi.org/10.18653/v1/2026.acl-long.1584) · [OpenAlex](https://openalex.org/W4417078035)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · [1](https://md.research.elamur.ai/free/en/papers/nghiem-2026-smarter-data-efficient-framework-improve/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary

SMARTER is a two-stage framework for toxicity detection in low-data settings. It aims to make LLM-based moderation both accurate and explainable. Two open-source models, COT-T5-XL and Llama-3.1-8B-Instruct, are first primed through LoRA-based SFT on HateCOT. They are then fine-tuned on small K-shot in-domain samples.

In Stage 1, each model expands its own training data. It is prompted to justify both the gold label and the wrong labels. This yields preferred and dispreferred explanations that feed preference optimization (DPO or KTO) without extra human annotation. In Stage 2, cross-model refinement trains one model on explanations the other model produced on held-out samples, so a weaker model can absorb a stronger one's reasoning.

On HateXplain, Latent Hate and Implicit Hate, the approach improves macro-F1 by up to roughly 13% over few-shot baselines. It also reaches most of full-training performance with a fraction of the labeled data. This makes transparent, self-hosted moderation feasible without large annotation budgets or commercial APIs.

## Key points

- **Tasks:** Three classification tasks come from two datasets.
  - HateXplain: 20,148 Gab/Twitter posts labeled Hate, Offensive or Normal.
  - Latent Hate: 19,112 tweets labeled Not Hate, Implicit Hate or Explicit Hate.
  - Implicit Hate: 4,153 tweets in 6 fine-grained categories, such as White Grievance and Irony.
- **Training sizes:** K-shot runs use K ∈ {16, 32, 64, 128, 256}. Synthetic explanations for incorrect labels serve as rejected samples for DPO and KTO.
- **Data efficiency:** The models achieve 86–100% of full-model performance using only 6–57% of the training data. They also give more stable results than 16-shot in-context learning with commercial APIs, whose scores vary by about ±0.10.
- **Cross-model refinement:** Cross-training at a K = 128 checkpoint lets T5 benefit from Llama's explanations. Mixing an encoder-decoder with a decoder-only architecture appears to help.
- **Limitations:**
  - Human evaluation covered only one dataset; raters preferred Llama's explanations on Normal posts by about 3:1.
  - The work is English-only.
  - Cross-training may slightly reduce consistency between explanations and labels.

*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

- [University of Maryland, College Park](https://md.research.elamur.ai/free/en/universities/university-of-maryland-college-park/) — University, College Park, United States · 132 works · 25,380 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,772 works · [🤝 Sociology and Political Science](https://md.research.elamur.ai/free/en/shelves/social-sciences/sociology-and-political-science/) · 577 works

🧭 Topics (1): [🛡️ Bias, fairness & safety](https://md.research.elamur.ai/free/en/topics/bias-fairness-safety/) · 517 works

## Cite

Cite as: Huy Nghiem, Advik Sachdeva and Hal Daumé Iii. 2026. SMARTER: A Data-efficient Framework to Improve Toxicity Detection with Explanation via Self-augmenting Large Language Models. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

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

```bibtex
@inproceedings{nghiem2026smarter,
  title = {{SMARTER: A Data-efficient Framework to Improve Toxicity Detection with Explanation via Self-augmenting Large Language Models}},
  author = {Huy Nghiem, Advik Sachdeva and Hal Daumé Iii},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1584},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1584}
}
```

- [← previous SmartAD: Capacity-Aligned Agent Distillation for Small Language Models](https://md.research.elamur.ai/free/en/papers/tang-2026-smartad-capacity-aligned-agent-distillation/)
- [next → SMTE at ATE-IT: Ensemble Term Extraction with Italian BERT, spaCy, and Vocabulary-Based Filtering](https://md.research.elamur.ai/free/en/papers/maule-2026-smte-ate-it-ensemble-term/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/17/)

## References

67 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Saad Almohaimeed, Saleh Almohaimeed, Damla Turgut, and Ladislau Bölöni. 2025. Towards generalizable generic harmful speech datasets for implicit hate speech detection. arXiv preprint arXiv:2506.16476. [find](https://scholar.google.com/scholar?q=Saad+Almohaimeed%2C+Saleh+Almohaimeed%2C+Damla+Turgut%2C+and+Ladislau+B%C3%B6l%C3%B6ni%2E+2025%2E+Towards+generalizable+generic+harmful+speech+datasets+for+implicit+hate+speech+detection%2E+arXiv+preprint+arXiv%3A2506%2E16476%2E)
2. Anjum and Rahul Katarya. 2024. Hate speech, toxicity detection in online social media: a recent survey of state of the art and opportunities. International Journal of Information Security, 23(1):577–608. [find](https://scholar.google.com/scholar?q=Anjum+and+Rahul+Katarya%2E+2024%2E+Hate+speech%2C+toxicity+detection+in+online+social+media%3A+a+recent+survey+of+state+of+the+art+and+opportunities%2E+International+Journal+of+Information+Security%2C+23%281%29%3A577%E2%80%936)
3. Dimosthenis Antypas and Jose Camacho-Collados. 2023. Robust hate speech detection in social media: A cross-dataset empirical evaluation. In The 7th Workshop on Online Abuse and Harms (WOAH), pages 231–242. [find](https://scholar.google.com/scholar?q=Dimosthenis+Antypas+and+Jose+Camacho%2DCollados%2E+2023%2E+Robust+hate+speech+detection+in+social+media%3A+A+cross%2Ddataset+empirical+evaluation%2E+In+The+7th+Workshop+on+Online+Abuse+and+Harms+%28WOAH%29%2C+pages+231)
4. Stephanie Alice Baker, Matthew Wade, and Michael James Walsh. 2020. \\<? covid19?\> the challenges of responding to misinformation during a pandemic: content moderation and the limitations of the concept of harm. Media International Australia, 177(1):103–107. [find](https://scholar.google.com/scholar?q=Stephanie+Alice+Baker%2C+Matthew+Wade%2C+and+Michael+James+Walsh%2E+2020%2E+%5C%3C%3F+covid19%3F%5C%3E+the+challenges+of+responding+to+misinformation+during+a+pandemic%3A+content+moderation+and+the+limitations+of+the+conce)
5. Dmitriy Bespalov, Sourav Bhabesh, Yi Xiang, Liutong Zhou, and Yanjun Qi. 2023. Towards building a robust toxicity predictor. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track), pages 581– 598. [find](https://scholar.google.com/scholar?q=Dmitriy+Bespalov%2C+Sourav+Bhabesh%2C+Yi+Xiang%2C+Liutong+Zhou%2C+and+Yanjun+Qi%2E+2023%2E+Towards+building+a+robust+toxicity+predictor%2E+In+Proceedings+of+the+61st+Annual+Meeting+of+the+Association+for+Computatio)
6. Agostina Calabrese, Leonardo Neves, Neil Shah, Maarten Bos, Björn Ross, Mirella Lapata, and Francesco Barbieri. 2024. Explainability and hate speech: Structured explanations make social media moderators faster. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistic… [find](https://scholar.google.com/scholar?q=Agostina+Calabrese%2C+Leonardo+Neves%2C+Neil+Shah%2C+Maarten+Bos%2C+Bj%C3%B6rn+Ross%2C+Mirella+Lapata%2C+and+Francesco+Barbieri%2E+2024%2E+Explainability+and+hate+speech%3A+Structured+explanations+make+social+media+moderato)
7. Yang Cao, Lovely-Frances Domingo, Sarah Gilbert, Michelle Mazurek, Katie Shilton, and Hal Daumé Iii. 2024. Toxicity detection is not all you need: Measuring the gaps to supporting volunteer content moderators through a user-centric method. In Proceedings of the 2024 Conference on Empirical Methods i… [find](https://scholar.google.com/scholar?q=Yang+Cao%2C+Lovely%2DFrances+Domingo%2C+Sarah+Gilbert%2C+Michelle+Mazurek%2C+Katie+Shilton%2C+and+Hal+Daum%C3%A9+Iii%2E+2024%2E+Toxicity+detection+is+not+all+you+need%3A+Measuring+the+gaps+to+supporting+volunteer+content+mo)
8. Tommaso Caselli, Valerio Basile, Jelena Mitrović, and Michael Granitzer. 2021. [Hatebert: Retraining bert for abusive language detection in english](https://md.research.elamur.ai/free/en/papers/caselli-2021-hatebert-retraining-bert-abusive-language/). In Proceedings of the 5th Workshop on Online Abuse and Harms (WOAH 2021), pages 17–25.
9. Sergio Andrés Castaño-Pulgarín, Natalia Suárez-Betancur, Luz Magnolia Tilano Vega, and Harvey Mauricio Herrera López. 2021. Internet, social media and online hate speech. systematic review. Aggression and violent behavior, 58:101608. [find](https://scholar.google.com/scholar?q=Sergio+Andr%C3%A9s+Casta%C3%B1o%2DPulgar%C3%ADn%2C+Natalia+Su%C3%A1rez%2DBetancur%2C+Luz+Magnolia+Tilano+Vega%2C+and+Harvey+Mauricio+Herrera+L%C3%B3pez%2E+2021%2E+Internet%2C+social+media+and+online+hate+speech%2E+systematic+review%2E+Aggression)
10. Peter Christen, David J Hand, and Nishadi Kirielle. 2023. A review of the f-measure: its history, properties, criticism, and alternatives. ACM Computing Surveys, 56(3):1–24. [find](https://scholar.google.com/scholar?q=Peter+Christen%2C+David+J+Hand%2C+and+Nishadi+Kirielle%2E+2023%2E+A+review+of+the+f%2Dmeasure%3A+its+history%2C+properties%2C+criticism%2C+and+alternatives%2E+ACM+Computing+Surveys%2C+56%283%29%3A1%E2%80%9324%2E)
11. Matteo Cinelli, Gianmarco De Francisci Morales, Alessandro Galeazzi, Walter Quattrociocchi, and Michele Starnini. 2021. The echo chamber effect on social media. Proceedings of the National Academy of Sciences, 118(9):e2023301118. [find](https://scholar.google.com/scholar?q=Matteo+Cinelli%2C+Gianmarco+De+Francisci+Morales%2C+Alessandro+Galeazzi%2C+Walter+Quattrociocchi%2C+and+Michele+Starnini%2E+2021%2E+The+echo+chamber+effect+on+social+media%2E+Proceedings+of+the+National+Academy+of+)
12. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies,…
13. Chiara Di Bonaventura, Lucia Siciliani, Pierpaolo Basile, Albert Merono Penuela, and Barbara McGillivray. 2024. Is explanation all you need? an expert survey on llm-generated explanations for abusive language detection. In Tenth Italian Conference on Computational Linguistics (CLiC-it 2024). [find](https://scholar.google.com/scholar?q=Chiara+Di+Bonaventura%2C+Lucia+Siciliani%2C+Pierpaolo+Basile%2C+Albert+Merono+Penuela%2C+and+Barbara+McGillivray%2E+2024%2E+Is+explanation+all+you+need%3F+an+expert+survey+on+llm%2Dgenerated+explanations+for+abusive+)
14. Thiago Dias Oliva. 2020. Content moderation technologies: Applying human rights standards to protect freedom of expression. Human Rights Law Review, 20(4):607–640. [find](https://scholar.google.com/scholar?q=Thiago+Dias+Oliva%2E+2020%2E+Content+moderation+technologies%3A+Applying+human+rights+standards+to+protect+freedom+of+expression%2E+Human+Rights+Law+Review%2C+20%284%29%3A607%E2%80%93640%2E)
15. Ashwin Geet d’Sa, Irina Illina, and Dominique Fohr. 2020. Bert and fasttext embeddings for automatic detection of toxic speech. In 2020 International Multi-Conference on:“Organization of Knowledge and Advanced Technologies”(OCTA), pages 1–5. IEEE. [find](https://scholar.google.com/scholar?q=Ashwin+Geet+d%E2%80%99Sa%2C+Irina+Illina%2C+and+Dominique+Fohr%2E+2020%2E+Bert+and+fasttext+embeddings+for+automatic+detection+of+toxic+speech%2E+In+2020+International+Multi%2DConference+on%3A%E2%80%9COrganization+of+Knowledge+and)
16. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv preprint arXiv:2407.21783.
17. Mai ElSherief, Caleb Ziems, David Muchlinski, Vaishnavi Anupindi, Jordyn Seybolt, Munmun De Choudhury, and Diyi Yang. 2021. [Latent hatred: A benchmark for understanding implicit hate speech](https://md.research.elamur.ai/free/en/papers/elsherief-2021-latent-hatred-benchmark-understanding-implicit/). In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 345–363.
18. Kawin Ethayarajh, Winnie Xu, Niklas Muennighoff, Dan Jurafsky, and Douwe Kiela. 2024. Kto: Model alignment as prospect theoretic optimization. arXiv preprint arXiv:2402.01306. [find](https://scholar.google.com/scholar?q=Kawin+Ethayarajh%2C+Winnie+Xu%2C+Niklas+Muennighoff%2C+Dan+Jurafsky%2C+and+Douwe+Kiela%2E+2024%2E+Kto%3A+Model+alignment+as+prospect+theoretic+optimization%2E+arXiv+preprint+arXiv%3A2402%2E01306%2E)
19. Paula Fortuna, Juan Soler-Company, and Leo Wanner. 2021. How well do hate speech, toxicity, abusive and offensive language classification models generalize across datasets? Information Processing & Management, 58(3):102524. [find](https://scholar.google.com/scholar?q=Paula+Fortuna%2C+Juan+Soler%2DCompany%2C+and+Leo+Wanner%2E+2021%2E+How+well+do+hate+speech%2C+toxicity%2C+abusive+and+offensive+language+classification+models+generalize+across+datasets%3F+Information+Processing+%26+Ma)
20. Giuseppe Gallipoli and Luca Cagliero. 2025. [It is not a piece of cake for gpt: Explaining textual entailment recognition in the presence of figurative language](https://md.research.elamur.ai/free/en/papers/gallipoli-2025-it-is-not-piece-cake/). In Proceedings of the 31st International Conference on Computational Linguistics, pages 9656–9674.
21. Seungone Kim, Se Joo, Doyoung Kim, Joel Jang, Seonghyeon Ye, Jamin Shin, and Minjoon Seo. 2023. [The cot collection: Improving zero-shot and few-shot learning of language models via chain-of-thought fine-tuning](https://md.research.elamur.ai/free/en/papers/kim-2023-cot-collection-improving-zero-shot/). In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing…
22. Katharine Gelber. 2021. Differentiating hate speech: a systemic discrimination approach. Critical Review of International Social and Political Philosophy. [find](https://scholar.google.com/scholar?q=Katharine+Gelber%2E+2021%2E+Differentiating+hate+speech%3A+a+systemic+discrimination+approach%2E+Critical+Review+of+International+Social+and+Political+Philosophy%2E)
23. Anastasia Kozyreva, Stefan M Herzog, Stephan Lewandowsky, Ralph Hertwig, Philipp Lorenz-Spreen, Mark Leiser, and Jason Reifler. 2023. Resolving content moderation dilemmas between free speech and harmful misinformation. Proceedings of the National Academy of Sciences, 120(7):e2210666120. [find](https://scholar.google.com/scholar?q=Anastasia+Kozyreva%2C+Stefan+M+Herzog%2C+Stephan+Lewandowsky%2C+Ralph+Hertwig%2C+Philipp+Lorenz%2DSpreen%2C+Mark+Leiser%2C+and+Jason+Reifler%2E+2023%2E+Resolving+content+moderation+dilemmas+between+free+speech+and+harm)
24. Feng Gu, Zongxia Li, Carlos Rafael Colon, Benjamin Evans, Ishani Mondal, and Jordan Lee Boyd-Graber. 2025. Large language models are effective human annotation assistants, but not good independent annotators. arXiv preprint arXiv:2503.06778. [find](https://scholar.google.com/scholar?q=Feng+Gu%2C+Zongxia+Li%2C+Carlos+Rafael+Colon%2C+Benjamin+Evans%2C+Ishani+Mondal%2C+and+Jordan+Lee+Boyd%2DGraber%2E+2025%2E+Large+language+models+are+effective+human+annotation+assistants%2C+but+not+good+independent+ann)
25. Tharindu Kumarage, Amrita Bhattacharjee, and Joshua Garland. 2024. Harnessing artificial intelligence to combat online hate: Exploring the challenges and opportunities of large language models in hate speech detection. arXiv preprint arXiv:2403.08035. — [arXiv:2403.08035](https://arxiv.org/abs/2403.08035) [find](https://scholar.google.com/scholar?q=Tharindu+Kumarage%2C+Amrita+Bhattacharjee%2C+and+Joshua+Garland%2E+2024%2E+Harnessing+artificial+intelligence+to+combat+online+hate%3A+Exploring+the+challenges+and+opportunities+of+large+language+models+in+hate)
26. Keyan Guo, Alexander Hu, Jaden Mu, Ziheng Shi, Ziming Zhao, Nishant Vishwamitra, and Hongxin Hu. 2023. An investigation of large language models for real-world hate speech detection. In 2023 International Conference on Machine Learning and Applications (ICMLA), pages 1568–1573. IEEE. [find](https://scholar.google.com/scholar?q=Keyan+Guo%2C+Alexander+Hu%2C+Jaden+Mu%2C+Ziheng+Shi%2C+Ziming+Zhao%2C+Nishant+Vishwamitra%2C+and+Hongxin+Hu%2E+2023%2E+An+investigation+of+large+language+models+for+real%2Dworld+hate+speech+detection%2E+In+2023+Internati)
27. Vivian Lai, Samuel Carton, Rajat Bhatnagar, Q Vera Liao, Yunfeng Zhang, and Chenhao Tan. 2022. Human-ai collaboration via conditional delegation: A case study of content moderation. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems, pages 1–18. [find](https://scholar.google.com/scholar?q=Vivian+Lai%2C+Samuel+Carton%2C+Rajat+Bhatnagar%2C+Q+Vera+Liao%2C+Yunfeng+Zhang%2C+and+Chenhao+Tan%2E+2022%2E+Human%2Dai+collaboration+via+conditional+delegation%3A+A+case+study+of+content+moderation%2E+In+Proceedings+of+)
28. Shibo Hao, Yi Gu, Haotian Luo, Tianyang Liu, Xiyan Shao, Xinyuan Wang, Shuhua Xie, Haodi Ma, Adithya Samavedhi, Qiyue Gao, et al. 2024. Llm reasoners: New evaluation, library, and analysis of step-by-step reasoning with large language models. In First Conference on Language Modeling. [find](https://scholar.google.com/scholar?q=Shibo+Hao%2C+Yi+Gu%2C+Haotian+Luo%2C+Tianyang+Liu%2C+Xiyan+Shao%2C+Xinyuan+Wang%2C+Shuhua+Xie%2C+Haodi+Ma%2C+Adithya+Samavedhi%2C+Qiyue+Gao%2C+et+al%2E+2024%2E+Llm+reasoners%3A+New+evaluation%2C+library%2C+and+analysis+of+step%2Dby%2D)
29. Zhuoyan Li, Hangxiao Zhu, Zhuoran Lu, and Ming Yin. 2023. [Synthetic data generation with large language models for text classification: Potential and limitations](https://md.research.elamur.ai/free/en/papers/li-2023-synthetic-data-generation-large-language/). In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 10443–10461.
30. Edward J Hu, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. 2021. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). In International Conference on Learning Representations.
31. Shayne Longpre, Le Hou, Tu Vu, Albert Webson, Hyung Won Chung, Yi Tay, Denny Zhou, Quoc V Le, Barret Zoph, Jason Wei, et al. 2023. [The flan collection: Designing data and methods for effective instruction tuning](https://md.research.elamur.ai/free/en/papers/longpre-2023-flan-collection-designing-data-methods/). In International Conference on Machine Learning, pages 22631–22648. PMLR.
32. Hyesun Hwang and Kee-Ok Kim. 2015. Social media as a tool for social movements: The effect of social media use and social capital on intention to participate in social movements. International Journal of Consumer Studies, 39(5):478–488. [find](https://scholar.google.com/scholar?q=Hyesun+Hwang+and+Kee%2DOk+Kim%2E+2015%2E+Social+media+as+a+tool+for+social+movements%3A+The+effect+of+social+media+use+and+social+capital+on+intention+to+participate+in+social+movements%2E+International+Journal)
33. Md Saroar Jahan, Mourad Oussalah, Djamila Romaissa Beddia, Nabil Arhab, et al. 2024. A comprehensive study on nlp data augmentation for hate speech detection: Legacy methods, bert, and llms. arXiv preprint arXiv:2404.00303. — [arXiv:2404.00303](https://arxiv.org/abs/2404.00303) [find](https://scholar.google.com/scholar?q=Md+Saroar+Jahan%2C+Mourad+Oussalah%2C+Djamila+Romaissa+Beddia%2C+Nabil+Arhab%2C+et+al%2E+2024%2E+A+comprehensive+study+on+nlp+data+augmentation+for+hate+speech+detection%3A+Legacy+methods%2C+bert%2C+and+llms%2E+arXiv+pre)
34. Daniel Kahneman and Amos Tversky. 2013. Prospect theory: An analysis of decision under risk. In Handbook of the fundamentals of financial decision making: Part I, pages 99–127. World Scientific. [find](https://scholar.google.com/scholar?q=Daniel+Kahneman+and+Amos+Tversky%2E+2013%2E+Prospect+theory%3A+An+analysis+of+decision+under+risk%2E+In+Handbook+of+the+fundamentals+of+financial+decision+making%3A+Part+I%2C+pages+99%E2%80%93127%2E+World+Scientific%2E)
35. Jaehoon Kim, Seungwan Jin, Sohyun Park, Someen Park, and Kyungsik Han. 2024. Label-aware hard negative sampling strategies with momentum contrastive learning for implicit hate speech detection. In Findings of the Association for Computational Linguistics ACL 2024, pages 16177–16188. [find](https://scholar.google.com/scholar?q=Jaehoon+Kim%2C+Seungwan+Jin%2C+Sohyun+Park%2C+Someen+Park%2C+and+Kyungsik+Han%2E+2024%2E+Label%2Daware+hard+negative+sampling+strategies+with+momentum+contrastive+learning+for+implicit+hate+speech+detection%2E+In+Fin)
36. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, et al. 2023. Self-refine: Iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 36:46534–46594. [find](https://scholar.google.com/scholar?q=Aman+Madaan%2C+Niket+Tandon%2C+Prakhar+Gupta%2C+Skyler+Hallinan%2C+Luyu+Gao%2C+Sarah+Wiegreffe%2C+Uri+Alon%2C+Nouha+Dziri%2C+Shrimai+Prabhumoye%2C+Yiming+Yang%2C+et+al%2E+2023%2E+Self%2Drefine%3A+Iterative+refinement+with+self%2Df)
37. Sarah Masud, Sahajpreet Singh, Viktor Hangya, Alexander Fraser, and Tanmoy Chakraborty. 2024. Hate personified: Investigating the role of llms in content moderation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 15847–15863. [find](https://scholar.google.com/scholar?q=Sarah+Masud%2C+Sahajpreet+Singh%2C+Viktor+Hangya%2C+Alexander+Fraser%2C+and+Tanmoy+Chakraborty%2E+2024%2E+Hate+personified%3A+Investigating+the+role+of+llms+in+content+moderation%2E+In+Proceedings+of+the+2024+Confere)
38. Binny Mathew, Punyajoy Saha, Seid Muhie Yimam, Chris Biemann, Pawan Goyal, and Animesh Mukherjee. 2021. Hatexplain: A benchmark dataset for explainable hate speech detection. In Proceedings of the AAAI conference on artificial intelligence, volume 35, pages 14867–14875. [find](https://scholar.google.com/scholar?q=Binny+Mathew%2C+Punyajoy+Saha%2C+Seid+Muhie+Yimam%2C+Chris+Biemann%2C+Pawan+Goyal%2C+and+Animesh+Mukherjee%2E+2021%2E+Hatexplain%3A+A+benchmark+dataset+for+explainable+hate+speech+detection%2E+In+Proceedings+of+the+AAA)
39. Vlad Andrei Negru, Robert Vacareanu, Camelia Lem-naru, Mihai Surdeanu, and Rodica Potolea. 2025. [find](https://scholar.google.com/scholar?q=Vlad+Andrei+Negru%2C+Robert+Vacareanu%2C+Camelia+Lem%2Dnaru%2C+Mihai+Surdeanu%2C+and+Rodica+Potolea%2E+2025%2E)
40. Morphnli: A stepwise approach to natural language inference using text morphing. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 6938–6953. [find](https://scholar.google.com/scholar?q=Morphnli%3A+A+stepwise+approach+to+natural+language+inference+using+text+morphing%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+NAACL+2025%2C+pages+6938%E2%80%936953%2E)
41. Huy Nghiem and Hal Daumé. 2024. HateCOT: An explanation-enhanced dataset for generalizable offensive speech detection via large language models. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 5938–5956, Miami, Florida, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Huy+Nghiem+and+Hal+Daum%C3%A9%2E+2024%2E+HateCOT%3A+An+explanation%2Denhanced+dataset+for+generalizable+offensive+speech+detection+via+large+language+models%2E+In+Findings+of+the+Association+for+Computational+Lingui)
42. Huy Nghiem, Umang Gupta, and Fred Morstatter. 2024. “define your terms”: Enhancing efficient offensive speech classification with definition. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1293–1309. [find](https://scholar.google.com/scholar?q=Huy+Nghiem%2C+Umang+Gupta%2C+and+Fred+Morstatter%2E+2024%2E+%E2%80%9Cdefine+your+terms%E2%80%9D%3A+Enhancing+efficient+offensive+speech+classification+with+definition%2E+In+Proceedings+of+the+18th+Conference+of+the+European+Chap)
43. Huy Nghiem, S. Panda, D. Khatwani, H. V. Nguyen, K. Kenthapadi, and H. D. Nguyen. 2025. Balancing safety and helpfulness in healthcare ai assistants through iterative preference alignment. In ML4H Symposium. ArXiv preprint arXiv:2512.04210. — [arXiv:2512.04210](https://arxiv.org/abs/2512.04210) [find](https://scholar.google.com/scholar?q=Huy+Nghiem%2C+S%2E+Panda%2C+D%2E+Khatwani%2C+H%2E+V%2E+Nguyen%2C+K%2E+Kenthapadi%2C+and+H%2E+D%2E+Nguyen%2E+2025%2E+Balancing+safety+and+helpfulness+in+healthcare+ai+assistants+through+iterative+preference+alignment%2E+In+ML4H+Sym)
44. Thu T Nguyen, Junaid S Merchant, Xiaohe Yue, Heran Mane, Hanxue Wei, Dina Huang, Krishik N Gowda, Katrina Makres, Crystal Najib, Huy T Nghiem, et al. 2024. A decade of tweets: Visualizing racial sentiments towards minoritized groups in the united states between 2011 and 2021. Epidemiology, 35(1):51–… [find](https://scholar.google.com/scholar?q=Thu+T+Nguyen%2C+Junaid+S+Merchant%2C+Xiaohe+Yue%2C+Heran+Mane%2C+Hanxue+Wei%2C+Dina+Huang%2C+Krishik+N+Gowda%2C+Katrina+Makres%2C+Crystal+Najib%2C+Huy+T+Nghiem%2C+et+al%2E+2024%2E+A+decade+of+tweets%3A+Visualizing+racial+senti)
45. Tin Nguyen, Jiannan Xu, Aayushi Roy, Hal Daumé III, and Marine Carpuat. 2023. Towards conceptualization of “fair explanation”: Disparate impacts of anti-asian hate speech explanations on content moderators. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pa… [find](https://scholar.google.com/scholar?q=Tin+Nguyen%2C+Jiannan+Xu%2C+Aayushi+Roy%2C+Hal+Daum%C3%A9+III%2C+and+Marine+Carpuat%2E+2023%2E+Towards+conceptualization+of+%E2%80%9Cfair+explanation%E2%80%9D%3A+Disparate+impacts+of+anti%2Dasian+hate+speech+explanations+on+content+moder)
46. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. [Training language models to follow instructions with human feedback](https://md.research.elamur.ai/free/en/papers/ouyang-2022-training-language-models-follow-instructions/). Advances in neural information processing systems, 35:27730–27744.
47. Melissa Kazemi Rad, Huy Nghiem, Andy Luo, Sahil Wadhwa, Mohammad Sorower, and Stephen Rawls. 2025. [Refining input guardrails: Enhancing llm-as-a-judge efficiency through chain-of-thought fine-tuning and alignment](https://md.research.elamur.ai/free/en/papers/rad-2025-refining-input-guardrails-enhancing-llm/). arXiv preprint arXiv:2501.13080.
48. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2024. [Direct preference optimization: Your language model is secretly a reward model](https://md.research.elamur.ai/free/en/papers/rafailov-2023-direct-preference-optimization-language-model/). Advances in Neural Information Processing Systems, 36.
49. Sarthak Roy, Ashish Harshvardhan, Animesh Mukherjee, and Punyajoy Saha. 2023. Probing llms for hate speech detection: strengths and vulnerabilities. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 6116–6128. [find](https://scholar.google.com/scholar?q=Sarthak+Roy%2C+Ashish+Harshvardhan%2C+Animesh+Mukherjee%2C+and+Punyajoy+Saha%2E+2023%2E+Probing+llms+for+hate+speech+detection%3A+strengths+and+vulnerabilities%2E+In+Findings+of+the+Association+for+Computational+Li)
50. Maarten Sap, Swabha Swayamdipta, Laura Vianna, Xuhui Zhou, Yejin Choi, and Noah A Smith. 2021. [Annotators with attitudes: How annotator beliefs and identities bias toxic language detection](https://md.research.elamur.ai/free/en/papers/sap-2022-annotators-attitudes-how-annotator-beliefs/). arXiv preprint arXiv:2111.07997.
51. Diptanu Sarkar, Marcos Zampieri, Tharindu Ranasinghe, and Alexander Ororbia. 2021. fbert: A neural transformer for identifying offensive content. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 1792–1798. [find](https://scholar.google.com/scholar?q=Diptanu+Sarkar%2C+Marcos+Zampieri%2C+Tharindu+Ranasinghe%2C+and+Alexander+Ororbia%2E+2021%2E+fbert%3A+A+neural+transformer+for+identifying+offensive+content%2E+In+Findings+of+the+Association+for+Computational+Lingu)
52. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, et al. 2024. [Deepseekmath: Pushing the limits of mathematical reasoning in open language models](https://md.research.elamur.ai/free/en/papers/shao-2024-deepseekmath-pushing-limits-mathematical-reasoning/). arXiv preprint arXiv:2402.03300.
53. Ruth Spence, Antonia Bifulco, Paula Bradbury, Elena Martellozzo, and Jeffrey DeMarco. 2023. The psychological impacts of content moderation on content moderators: A qualitative study. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 17(4). [find](https://scholar.google.com/scholar?q=Ruth+Spence%2C+Antonia+Bifulco%2C+Paula+Bradbury%2C+Elena+Martellozzo%2C+and+Jeffrey+DeMarco%2E+2023%2E+The+psychological+impacts+of+content+moderation+on+content+moderators%3A+A+qualitative+study%2E+Cyberpsychology%3A)
54. Pabba Sumanth, Syed Samiuddin, K Jamal, Srikanth Domakonda, and Pathi Shivani. 2022. Toxic speech classification using machine learning algorithms. In 2022 International Conference on Electronic Systems and Intelligent Computing (ICESIC), pages 257–263. IEEE. [find](https://scholar.google.com/scholar?q=Pabba+Sumanth%2C+Syed+Samiuddin%2C+K+Jamal%2C+Srikanth+Domakonda%2C+and+Pathi+Shivani%2E+2022%2E+Toxic+speech+classification+using+machine+learning+algorithms%2E+In+2022+International+Conference+on+Electronic+Syste)
55. Ruixiang Tang, Xiaotian Han, Xiaoqian Jiang, and Xia Hu. 2023. Does synthetic data generation of llms help clinical text mining? arXiv preprint arXiv:2303.04360. — [arXiv:2303.04360](https://arxiv.org/abs/2303.04360) [find](https://scholar.google.com/scholar?q=Ruixiang+Tang%2C+Xiaotian+Han%2C+Xiaoqian+Jiang%2C+and+Xia+Hu%2E+2023%2E+Does+synthetic+data+generation+of+llms+help+clinical+text+mining%3F+arXiv+preprint+arXiv%3A2303%2E04360%2E+%E2%80%94+%5BarXiv%3A2303%2E04360%5D%28https%3A%2F%2Farxiv%2Eorg)
56. Alice Tontodimamma, Eugenia Nissi, Annalina Sarra, and Lara Fontanella. 2021. Thirty years of research into hate speech: topics of interest and their evolution. Scientometrics, 126:157–179. [find](https://scholar.google.com/scholar?q=Alice+Tontodimamma%2C+Eugenia+Nissi%2C+Annalina+Sarra%2C+and+Lara+Fontanella%2E+2021%2E+Thirty+years+of+research+into+hate+speech%3A+topics+of+interest+and+their+evolution%2E+Scientometrics%2C+126%3A157%E2%80%93179%2E)
57. Cagri Toraman, Furkan Şahinuç, and Eyup Halit Yilmaz. 2022. Large-scale hate speech detection with cross-domain transfer. arXiv preprint arXiv:2203.01111. [find](https://scholar.google.com/scholar?q=Cagri+Toraman%2C+Furkan+S%CC%A7ahinu%C3%A7%2C+and+Eyup+Halit+Yilmaz%2E+2022%2E+Large%2Dscale+hate+speech+detection+with+cross%2Ddomain+transfer%2E+arXiv+preprint+arXiv%3A2203%2E01111%2E)
58. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2022. Self-instruct: Aligning language models with self-generated instructions. arXiv preprint arXiv:2212.10560. [find](https://scholar.google.com/scholar?q=Yizhong+Wang%2C+Yeganeh+Kordi%2C+Swaroop+Mishra%2C+Alisa+Liu%2C+Noah+A+Smith%2C+Daniel+Khashabi%2C+and+Hannaneh+Hajishirzi%2E+2022%2E+Self%2Dinstruct%3A+Aligning+language+models+with+self%2Dgenerated+instructions%2E+arXiv+pr)
59. Yulong Wang, Hong Li, and Ni Wei. 2025. Sahsd: Enhancing hate speech detection in llm-powered web applications via sentiment analysis and few-shot learning. In Proceedings of the ACM on Web Conference 2025, pages 3014–3025. [find](https://scholar.google.com/scholar?q=Yulong+Wang%2C+Hong+Li%2C+and+Ni+Wei%2E+2025%2E+Sahsd%3A+Enhancing+hate+speech+detection+in+llm%2Dpowered+web+applications+via+sentiment+analysis+and+few%2Dshot+learning%2E+In+Proceedings+of+the+ACM+on+Web+Conference)
60. Benjamin Warner, Antoine Chaffin, Benjamin Clavié, Orion Weller, Oskar Hallström, Said Taghadouini, Alexis Gallagher, Raja Biswas, Faisal Ladhak, Tom Aarsen, et al. 2024. [Smarter, better, faster, longer: A modern bidirectional encoder for fast, memory efficient, and long context finetuning and infer…](https://md.research.elamur.ai/free/en/papers/warner-2024-smarter-better-faster-longer-modern/)
61. Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. 2021. [Finetuned language models are zero-shot learners](https://md.research.elamur.ai/free/en/papers/wei-2021-finetuned-language-models-zero-shot/). arXiv preprint arXiv:2109.01652.
62. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. [Chain-of-thought prompting elicits reasoning in large language models](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). Advances in neural information processing systems, 35:24824–24837.
63. Steven Windisch, Susann Wiedlitzka, Ajima Olaghere, and Elizabeth Jenaway. 2022. Online interventions for reducing hate speech and cyberhate: A systematic review. Campbell systematic reviews, 18(2):e1243. [find](https://scholar.google.com/scholar?q=Steven+Windisch%2C+Susann+Wiedlitzka%2C+Ajima+Olaghere%2C+and+Elizabeth+Jenaway%2E+2022%2E+Online+interventions+for+reducing+hate+speech+and+cyberhate%3A+A+systematic+review%2E+Campbell+systematic+reviews%2C+18%282%29%3Ae1)
64. Yongjin Yang, Joonkee Kim, Yujin Kim, Namgyu Ho, James Thorne, and Se-Young Yun. 2023. Hare: Explainable hate speech detection with step-by-step reasoning. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 5490–5505. [find](https://scholar.google.com/scholar?q=Yongjin+Yang%2C+Joonkee+Kim%2C+Yujin+Kim%2C+Namgyu+Ho%2C+James+Thorne%2C+and+Se%2DYoung+Yun%2E+2023%2E+Hare%3A+Explainable+hate+speech+detection+with+step%2Dby%2Dstep+reasoning%2E+In+Findings+of+the+Association+for+Computati)
65. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2024. Tree of thoughts: Deliberate problem solving with large language models. Advances in Neural Information Processing Systems, 36. [find](https://scholar.google.com/scholar?q=Shunyu+Yao%2C+Dian+Yu%2C+Jeffrey+Zhao%2C+Izhak+Shafran%2C+Tom+Griffiths%2C+Yuan+Cao%2C+and+Karthik+Narasimhan%2E+2024%2E+Tree+of+thoughts%3A+Deliberate+problem+solving+with+large+language+models%2E+Advances+in+Neural+Inf)
66. Zhixue Zhao, Ziqi Zhang, and Frank Hopfgartner. 2021. A comparative study of using pre-trained language models for toxic comment classification. In Companion Proceedings of the Web Conference 2021, pages 500–507. [find](https://scholar.google.com/scholar?q=Zhixue+Zhao%2C+Ziqi+Zhang%2C+and+Frank+Hopfgartner%2E+2021%2E+A+comparative+study+of+using+pre%2Dtrained+language+models+for+toxic+comment+classification%2E+In+Companion+Proceedings+of+the+Web+Conference+2021%2C+pa)
67. Adam Zweiger, Jyothish Pari, Han Guo, Ekin Akyürek, Yoon Kim, and Pulkit Agrawal. 2025. Self-adapting language models. arXiv preprint arXiv:2506.10943. — [arXiv:2506.10943](https://arxiv.org/abs/2506.10943) [find](https://scholar.google.com/scholar?q=Adam+Zweiger%2C+Jyothish+Pari%2C+Han+Guo%2C+Ekin+Aky%C3%BCrek%2C+Yoon+Kim%2C+and+Pulkit+Agrawal%2E+2025%2E+Self%2Dadapting+language+models%2E+arXiv+preprint+arXiv%3A2506%2E10943%2E+%E2%80%94+%5BarXiv%3A2506%2E10943%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2506%2E)

*Updated 1 November 2026*