---
title: "HatePrototypes — Proskurina 2026"
description: "Hate speech classifiers are usually adapted to new kinds of hateful content through further fine-tuning."
authors: ["Irina Proskurina", "Marc-Antoine Carpentier", "Julien Velcin"]
year: 2026
venue: "Proceedings of the Fifteenth Language Resources and Evaluation Conference"
source: "https://aclanthology.org/2026.lrec-1.343/"
doi: "10.63317/3opu4zq9p6pc"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/proskurina-2026-hateprototypes-interpretable-transferable-representations-implicit/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/proskurina-2026-hateprototypes-interpretable-transferable-representations-implicit/)

[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/5/) / HatePrototypes: Interpretable and Transferable Representations for Implicit and Explicit Hate Speech Detection

# HatePrototypes: Interpretable and Transferable Representations for Implicit and Explicit Hate Speech Detection

**Authors:** Irina Proskurina, Marc-Antoine Carpentier, Julien Velcin · **Year:** 2026 · **Venue:** Proceedings of the Fifteenth Language Resources and Evaluation Conference · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.lrec-1.343/) · [PDF](https://aclanthology.org/2026.lrec-1.343.pdf) · [DOI 10.63317/3opu4zq9p6pc](https://doi.org/10.63317/3opu4zq9p6pc) · [OpenAlex](https://openalex.org/W4416157538)

**Citations:** 0 citations (openalex-snapshot · 2026-10-14)

## Summary
Hate speech classifiers are usually adapted to new kinds of hateful content through further fine-tuning. Yet common benchmarks centre on overt attacks on protected groups, while veiled hostility, such as coded language, demeaning comparisons or insinuation, is harder to detect and transfers poorly. The authors test whether this repeated retraining is needed. They introduce HatePrototypes: ℓ2-normalised mean embeddings of the hate and non-hate classes, taken from models tuned for hate detection or safety moderation. A new text gets the label of the centroid it is most similar to (cosine similarity). Building prototypes at every layer also enables a parameter-free early-exit rule: inference stops once the similarity gap between the two classes reaches a threshold δ.

Experiments use BERT-base and OPT-125M fine-tuned on two implicit benchmarks (IHC, SBIC) and two explicit ones (OLID, HateXplain), plus two guardrail models. Prototypes built from only 50 examples per class allow transfer between implicit and explicit tasks and can be swapped across benchmarks. Prototype-guided early exiting works for both hate types. The approach offers a cheap, interpretable alternative to retraining, and the code and prototypes are publicly released.

## Key points
- **Sentence representations:** BERT-base-cased (109M parameters) uses the [CLS] hidden state, and OPT-125M uses the last non-padding token. Both models have 12 layers, 12 heads and a hidden size of 768. The guardrail models are Llama-Guard-3-1B and BLOOMZ-Guardrail-3B.
- **Benchmark sizes (train/test) and hate share:**
  - SBIC: 35,424/4,691, 68.32% hate
  - IHC: 14,273/3,059, 34.82% hate
  - OLID: 9,268/860, 31.21% hate
  - HateXplain: 15,379/1,924, 59.36% hate
  - All four come mostly from Twitter, and SBIC and HateXplain also include Gab posts.
- **Fine-tuning setup:** 3 epochs, learning rate 1×10⁻⁵, batch size 64. Weighted cross-entropy handles class imbalance. Results are averaged over 10 random seeds on a single A100 GPU.
- **Early exiting:** a forward pass ends at the first layer where the top-1 vs. top-2 prototype similarity margin is at least δ. If no layer meets this, the final layer decides. No extra classification heads are trained.
- **Main finding:** prototypes from 50 examples per class transfer between implicit and explicit hate tasks, consistently for both model families. The authors present this as the first prototype-based analysis of implicit/explicit transfer and of early exiting for hate detection.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

## Cite

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

```bibtex
@inproceedings{proskurina2026hateprototypes,
  title = {{HatePrototypes: Interpretable and Transferable Representations for Implicit and Explicit Hate Speech Detection}},
  author = {Proskurina, Irina and Carpentier, Marc-Antoine and Velcin, Julien},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/3opu4zq9p6pc},
  url = {https://doi.org/10.63317/3opu4zq9p6pc}
}
```

- [← previous Harf-Speech: A Clinically Aligned Framework for Arabic Phoneme-Level Speech Assessment](https://md.research.elamur.ai/free/en/papers/azad-2026-harf-speech-clinically-aligned-framework/)
- [next → HCMUS_The Fangs at AbjadStyleTransfer Shared Task: Learning to Query Style, Contrastive Representations for Zero-Shot Arabic Authorship Style Transfer](https://md.research.elamur.ai/free/en/papers/sy-2026-hcmus-fangs-abjadstyletransfer-shared-task/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/5/)

## References

48 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Hyeseon Ahn, Youngwook Kim, Jungin Kim, and Yo-Sub Han. 2024. SharedCon: Implicit hate speech detection using shared semantics. In Findings of the Association for Computational Linguistics: ACL 2024, pages 10444–10455, Bangkok, Thailand. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Hyeseon+Ahn%2C+Youngwook+Kim%2C+Jungin+Kim%2C+and+Yo%2DSub+Han%2E+2024%2E+SharedCon%3A+Implicit+hate+speech+detection+using+shared+semantics%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+ACL+2024%2C+)
2. Abdullah Albanyan, Ahmed Hassan, and Eduardo Blanco. 2023. Finding authentic counterhate arguments: A case study with public figures. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 13862–13876, Singapore. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Abdullah+Albanyan%2C+Ahmed+Hassan%2C+and+Eduardo+Blanco%2E+2023%2E+Finding+authentic+counterhate+arguments%3A+A+case+study+with+public+figures%2E+In+Proceedings+of+the+2023+Conference+on+Empirical+Methods+in+Natu)
3. A. Arango, J. Pérez, and B. Poblete. 2019. Hate speech detection is not as easy as you may think: A closer look at model evaluation. In Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 45–54. ACM. [find](https://scholar.google.com/scholar?q=A%2E+Arango%2C+J%2E+P%C3%A9rez%2C+and+B%2E+Poblete%2E+2019%2E+Hate+speech+detection+is+not+as+easy+as+you+may+think%3A+A+closer+look+at+model+evaluation%2E+In+Proceedings+of+the+42nd+International+ACM+SIGIR+Conference+on+Re)
4. Sangmin Bae, Jongwoo Ko, Hwanjun Song, and Se-Young Yun. 2023. Fast and robust early-exiting framework for autoregressive language models with synchronized parallel decoding. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 5910–5924, Singapore. Associ… [find](https://scholar.google.com/scholar?q=Sangmin+Bae%2C+Jongwoo+Ko%2C+Hwanjun+Song%2C+and+Se%2DYoung+Yun%2E+2023%2E+Fast+and+robust+early%2Dexiting+framework+for+autoregressive+language+models+with+synchronized+parallel+decoding%2E+In+Proceedings+of+the+202)
5. David Dale, Anton Voronov, Daryna Dementieva, Varvara Logacheva, Olga Kozlova, Nikita Semenov, and Alexander Panchenko. 2021. Text detoxification using large pre-trained neural models. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 7979–7996, Online… [find](https://scholar.google.com/scholar?q=David+Dale%2C+Anton+Voronov%2C+Daryna+Dementieva%2C+Varvara+Logacheva%2C+Olga+Kozlova%2C+Nikita+Semenov%2C+and+Alexander+Panchenko%2E+2021%2E+Text+detoxification+using+large+pre%2Dtrained+neural+models%2E+In+Proceedings+)
6. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,… [find](https://scholar.google.com/scholar?q=Jacob+Devlin%2C+Ming%2DWei+Chang%2C+Kenton+Lee%2C+and+Kristina+Toutanova%2E+2019%2E+BERT%3A+Pre%2Dtraining+of+deep+bidirectional+transformers+for+language+understanding%2E+In+Proceedings+of+the+2019+Conference+of+the+N)
7. Ángel Díaz and Laura Hecht-Felella. 2021. Double standards in social media content moderation. Brennan Center for Justice at New York University School of Law, pages 1–23. [find](https://scholar.google.com/scholar?q=%C3%81ngel+D%C3%ADaz+and+Laura+Hecht%2DFelella%2E+2021%2E+Double+standards+in+social+media+content+moderation%2E+Brennan+Center+for+Justice+at+New+York+University+School+of+Law%2C+pages+1%E2%80%9323%2E)
8. Mostafa Elhoushi, Akshat Shrivastava, Diana Liskovich, Basil Hosmer, Bram Wasti, Liangzhen Lai, Anas Mahmoud, Bilge Acun, Saurabh Agarwal, Ahmed Roman, Ahmed Aly, Beidi Chen, and Carole-Jean Wu. 2024. LayerSkip: Enabling early exit inference and self-speculative decoding. In Proceedings of the 62nd… [find](https://scholar.google.com/scholar?q=Mostafa+Elhoushi%2C+Akshat+Shrivastava%2C+Diana+Liskovich%2C+Basil+Hosmer%2C+Bram+Wasti%2C+Liangzhen+Lai%2C+Anas+Mahmoud%2C+Bilge+Acun%2C+Saurabh+Agarwal%2C+Ahmed+Roman%2C+Ahmed+Aly%2C+Beidi+Chen%2C+and+Carole%2DJean+Wu%2E+2024%2E) [our card](https://md.research.elamur.ai/free/en/papers/elhoushi-2024-layerskip-enabling-early-exit-inference/)
9. Alperen Görmez, Venkat R. Dasari, and Erdem Koyuncu. 2022. E2cm: Early exit via class means for efficient supervised and unsupervised learning. In 2022 International Joint Conference on Neural Networks (IJCNN), pages 1–8. [find](https://scholar.google.com/scholar?q=Alperen+G%C3%B6rmez%2C+Venkat+R%2E+Dasari%2C+and+Erdem+Koyuncu%2E+2022%2E+E2cm%3A+Early+exit+via+class+means+for+efficient+supervised+and+unsupervised+learning%2E+In+2022+International+Joint+Conference+on+Neural+Network)
10. Thomas Hartvigsen, Saadia Gabriel, Hamid Palangi, Maarten Sap, Dipankar Ray, and Ece Kamar. 2022. ToxiGen: A large-scale machine-generated dataset for adversarial and implicit hate speech detection. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1:… [find](https://scholar.google.com/scholar?q=Thomas+Hartvigsen%2C+Saadia+Gabriel%2C+Hamid+Palangi%2C+Maarten+Sap%2C+Dipankar+Ray%2C+and+Ece+Kamar%2E+2022%2E+ToxiGen%3A+A+large%2Dscale+machine%2Dgenerated+dataset+for+adversarial+and+implicit+hate+speech+detection%2E+I) [our card](https://md.research.elamur.ai/free/en/papers/hartvigsen-2022-toxigen-large-scale-machine-generated/)
11. Jianing He, Qi Zhang, Weiping Ding, Duoqian Miao, Jun Zhao, Liang Hu, and Longbing Cao. 2024. De 3 -bert: Distance-enhanced early exiting for bert based on prototypical networks. [find](https://scholar.google.com/scholar?q=Jianing+He%2C+Qi+Zhang%2C+Weiping+Ding%2C+Duoqian+Miao%2C+Jun+Zhao%2C+Liang+Hu%2C+and+Longbing+Cao%2E+2024%2E+De+3+%2Dbert%3A+Distance%2Denhanced+early+exiting+for+bert+based+on+prototypical+networks%2E)
12. Hakan Inan, Kartikeya Upasani, Jianfeng Chi, Rashi Rungta, Krithika Iyer, Yuning Mao, Michael Tontchev, Qing Hu, Brian Fuller, Davide Testuggine, and Madian Khabsa. 2023. Llama guard: Llm-based input-output safeguard for human-ai conversations. [find](https://scholar.google.com/scholar?q=Hakan+Inan%2C+Kartikeya+Upasani%2C+Jianfeng+Chi%2C+Rashi+Rungta%2C+Krithika+Iyer%2C+Yuning+Mao%2C+Michael+Tontchev%2C+Qing+Hu%2C+Brian+Fuller%2C+Davide+Testuggine%2C+and+Madian+Khabsa%2E+2023%2E+Llama+guard%3A+Llm%2Dbased+input%2D) [our card](https://md.research.elamur.ai/free/en/papers/inan-2023-llama-guard-llm-based-input/)
13. Amir Reza Jafari, Guanlin Li, Praboda Rajapaksha, Reza Farahbakhsh, and Noel Crespi. 2023. Fine-grained emotions influence on implicit hate speech detection. IEEE Access, 11:105330– 105343. [find](https://scholar.google.com/scholar?q=Amir+Reza+Jafari%2C+Guanlin+Li%2C+Praboda+Rajapaksha%2C+Reza+Farahbakhsh%2C+and+Noel+Crespi%2E+2023%2E+Fine%2Dgrained+emotions+influence+on+implicit+hate+speech+detection%2E+IEEE+Access%2C+11%3A105330%E2%80%93+105343%2E)
14. Yiping Jin, Leo Wanner, and Alexander Shvets. 2024. GPT-HateCheck: Can LLMs write better functional tests for hate speech detection? In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 7867–7885, Torino,… [find](https://scholar.google.com/scholar?q=Yiping+Jin%2C+Leo+Wanner%2C+and+Alexander+Shvets%2E+2024%2E+GPT%2DHateCheck%3A+Can+LLMs+write+better+functional+tests+for+hate+speech+detection%3F+In+Proceedings+of+the+2024+Joint+International+Conference+on+Comput)
15. Urja Khurana, Ivar Vermeulen, Eric Nalisnick, Marloes Van Noorloos, and Antske Fokkens. 2022. Hate speech criteria: A modular approach to task-specific hate speech definitions. In Proceedings of the Sixth Workshop on Online Abuse and Harms (WOAH), pages 176–191, Seattle, Washington (Hybrid). Associa… [find](https://scholar.google.com/scholar?q=Urja+Khurana%2C+Ivar+Vermeulen%2C+Eric+Nalisnick%2C+Marloes+Van+Noorloos%2C+and+Antske+Fokkens%2E+2022%2E+Hate+speech+criteria%3A+A+modular+approach+to+task%2Dspecific+hate+speech+definitions%2E+In+Proceedings+of+the+S)
16. Youngwook Kim, Shinwoo Park, and Yo-Sub Han. 2022. Generalizable implicit hate speech detection using contrastive learning. In Proceedings of the 29th International Conference on Computational Linguistics, pages 6667–6679, Gyeongju, Republic of Korea. International Committee on Computational Linguis… [find](https://scholar.google.com/scholar?q=Youngwook+Kim%2C+Shinwoo+Park%2C+and+Yo%2DSub+Han%2E+2022%2E+Generalizable+implicit+hate+speech+detection+using+contrastive+learning%2E+In+Proceedings+of+the+29th+International+Conference+on+Computational+Linguis)
17. Alyssa Lees, Vinh Q. Tran, Yi Tay, Jeffrey Sorensen, Jai Gupta, Donald Metzler, and Lucy Vasserman. 2022. A new generation of perspective api: Efficient multilingual character-level transformers. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD ’22, page 3… [find](https://scholar.google.com/scholar?q=Alyssa+Lees%2C+Vinh+Q%2E+Tran%2C+Yi+Tay%2C+Jeffrey+Sorensen%2C+Jai+Gupta%2C+Donald+Metzler%2C+and+Lucy+Vasserman%2E+2022%2E+A+new+generation+of+perspective+api%3A+Efficient+multilingual+character%2Dlevel+transformers%2E+In+P)
18. João Leite, Carolina Scarton, and Diego Silva. 2023. Noisy self-training with data augmentations for offensive and hate speech detection tasks. In Proceedings of the 14th International Conference on Recent Advances in Natural Language Processing, pages 631–640, Varna, Bulgaria. INCOMA Ltd., Shoumen,… [find](https://scholar.google.com/scholar?q=Jo%C3%A3o+Leite%2C+Carolina+Scarton%2C+and+Diego+Silva%2E+2023%2E+Noisy+self%2Dtraining+with+data+augmentations+for+offensive+and+hate+speech+detection+tasks%2E+In+Proceedings+of+the+14th+International+Conference+on+R)
19. Weijie Liu, Peng Zhou, Zhiruo Wang, Zhe Zhao, Haotang Deng, and Qi Ju. 2020. FastBERT: a self-distilling BERT with adaptive inference time. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6035–6044, Online. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Weijie+Liu%2C+Peng+Zhou%2C+Zhiruo+Wang%2C+Zhe+Zhao%2C+Haotang+Deng%2C+and+Qi+Ju%2E+2020%2E+FastBERT%3A+a+self%2Ddistilling+BERT+with+adaptive+inference+time%2E+In+Proceedings+of+the+58th+Annual+Meeting+of+the+Association) [our card](https://md.research.elamur.ai/free/en/papers/liu-2020-fastbert-self-distilling-bert-adaptive/)
20. Xiangyang Liu, Tianxiang Sun, Junliang He, Jiawen Wu, Lingling Wu, Xinyu Zhang, Hao Jiang, Zhao Cao, Xuanjing Huang, and Xipeng Qiu. 2022. Towards efficient NLP: A standard evaluation and a strong baseline. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Co… [find](https://scholar.google.com/scholar?q=Xiangyang+Liu%2C+Tianxiang+Sun%2C+Junliang+He%2C+Jiawen+Wu%2C+Lingling+Wu%2C+Xinyu+Zhang%2C+Hao+Jiang%2C+Zhao+Cao%2C+Xuanjing+Huang%2C+and+Xipeng+Qiu%2E+2022%2E+Towards+efficient+NLP%3A+A+standard+evaluation+and+a+strong+bas)
21. Sean MacAvaney, Hao-Ren Yao, Eugene Yang, Katina Russell, Nazli Goharian, and Ophir Frieder. 2019. Hate speech detection: Challenges and solutions. PLOS ONE, 14(8):1–16. [find](https://scholar.google.com/scholar?q=Sean+MacAvaney%2C+Hao%2DRen+Yao%2C+Eugene+Yang%2C+Katina+Russell%2C+Nazli+Goharian%2C+and+Ophir+Frieder%2E+2019%2E+Hate+speech+detection%3A+Challenges+and+solutions%2E+PLOS+ONE%2C+14%288%29%3A1%E2%80%9316%2E)
22. Karsten Müller and Carlo Schwarz. 2020. Fanning the flames of hate: Social media and hate crime. Journal of the European Economic Association, 19(4):2131–2167. [find](https://scholar.google.com/scholar?q=Karsten+M%C3%BCller+and+Carlo+Schwarz%2E+2020%2E+Fanning+the+flames+of+hate%3A+Social+media+and+hate+crime%2E+Journal+of+the+European+Economic+Association%2C+19%284%29%3A2131%E2%80%932167%2E)
23. Moin Nadeem, Anna Bethke, and Siva Reddy. 2021. StereoSet: Measuring stereotypical bias in pre-trained language models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Lon… [find](https://scholar.google.com/scholar?q=Moin+Nadeem%2C+Anna+Bethke%2C+and+Siva+Reddy%2E+2021%2E+StereoSet%3A+Measuring+stereotypical+bias+in+pre%2Dtrained+language+models%2E+In+Proceedings+of+the+59th+Annual+Meeting+of+the+Association+for+Computational+L)
24. Nikita Nangia, Clara Vania, Rasika Bhalerao, and Samuel R. Bowman. 2020. CrowS-pairs: A challenge dataset for measuring social biases in masked language models. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1953–1967, Online. Association for… [find](https://scholar.google.com/scholar?q=Nikita+Nangia%2C+Clara+Vania%2C+Rasika+Bhalerao%2C+and+Samuel+R%2E+Bowman%2E+2020%2E+CrowS%2Dpairs%3A+A+challenge+dataset+for+measuring+social+biases+in+masked+language+models%2E+In+Proceedings+of+the+2020+Conference+o) [our card](https://md.research.elamur.ai/free/en/papers/nangia-2020-crows-pairs-challenge-dataset-measuring/)
25. Nicolás Benjamín Ocampo, Ekaterina Sviridova, Elena Cabrio, and Serena Villata. 2023. An in-depth analysis of implicit and subtle hate speech messages. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 1997– 2013, Dubrovnik, Croatia… [find](https://scholar.google.com/scholar?q=Nicol%C3%A1s+Benjam%C3%ADn+Ocampo%2C+Ekaterina+Sviridova%2C+Elena+Cabrio%2C+and+Serena+Villata%2E+2023%2E+An+in%2Ddepth+analysis+of+implicit+and+subtle+hate+speech+messages%2E+In+Proceedings+of+the+17th+Conference+of+the+Eur)
26. Yuto Oikawa, Yuki Nakayama, and Koji Murakami. 2022. A stacking-based efficient method for toxic language detection on live streaming chat. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 571–578, Abu Dhabi, UAE. Association for Comput… [find](https://scholar.google.com/scholar?q=Yuto+Oikawa%2C+Yuki+Nakayama%2C+and+Koji+Murakami%2E+2022%2E+A+stacking%2Dbased+efficient+method+for+toxic+language+detection+on+live+streaming+chat%2E+In+Proceedings+of+the+2022+Conference+on+Empirical+Methods+i)
27. Pia Pachinger, Allan Hanbury, Julia Neidhardt, and Anna Planitzer. 2023. Toward disambiguating the definitions of abusive, offensive, toxic, and uncivil comments. In Proceedings of the First Workshop on Cross-Cultural Considerations in NLP (C3NLP), pages 107–113, Dubrovnik, Croatia. Association for… [find](https://scholar.google.com/scholar?q=Pia+Pachinger%2C+Allan+Hanbury%2C+Julia+Neidhardt%2C+and+Anna+Planitzer%2E+2023%2E+Toward+disambiguating+the+definitions+of+abusive%2C+offensive%2C+toxic%2C+and+uncivil+comments%2E+In+Proceedings+of+the+First+Workshop+)
28. Haseena Rahmath P, Vishal Srivastava, Kuldeep Chaurasia, Roberto G. Pacheco, and Rodrigo S. Couto. 2024. Early-exit deep neural network - a comprehensive survey. ACM Comput. Surv., 57(3). [find](https://scholar.google.com/scholar?q=Haseena+Rahmath+P%2C+Vishal+Srivastava%2C+Kuldeep+Chaurasia%2C+Roberto+G%2E+Pacheco%2C+and+Rodrigo+S%2E+Couto%2E+2024%2E+Early%2Dexit+deep+neural+network+%2D+a+comprehensive+survey%2E+ACM+Comput%2E+Surv%2E%2C+57%283%29%2E)
29. Paul Röttger, Bertie Vidgen, Dong Nguyen, Zeerak Waseem, Helen Margetts, and Janet Pierrehumbert. 2021. Hatecheck: Functional tests for hate speech detection models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference… [find](https://scholar.google.com/scholar?q=Paul+R%C3%B6ttger%2C+Bertie+Vidgen%2C+Dong+Nguyen%2C+Zeerak+Waseem%2C+Helen+Margetts%2C+and+Janet+Pierrehumbert%2E+2021%2E+Hatecheck%3A+Functional+tests+for+hate+speech+detection+models%2E+In+Proceedings+of+the+59th+Annual+) [our card](https://md.research.elamur.ai/free/en/papers/rottger-2021-hatecheck-functional-tests-hate-speech/)
30. Anna Schmidt and Michael Wiegand. 2017. A survey on hate speech detection using natural language processing. In Proceedings of the Fifth International Workshop on Natural Language Processing for Social Media, pages 1–10, Valencia, Spain. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Anna+Schmidt+and+Michael+Wiegand%2E+2017%2E+A+survey+on+hate+speech+detection+using+natural+language+processing%2E+In+Proceedings+of+the+Fifth+International+Workshop+on+Natural+Language+Processing+for+Socia) [our card](https://md.research.elamur.ai/free/en/papers/schmidt-2017-survey-hate-speech-detection-natural/)
31. Jake Snell, Kevin Swersky, and Richard Zemel. 2017. Prototypical networks for few-shot learning. In Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc. [find](https://scholar.google.com/scholar?q=Jake+Snell%2C+Kevin+Swersky%2C+and+Richard+Zemel%2E+2017%2E+Prototypical+networks+for+few%2Dshot+learning%2E+In+Advances+in+Neural+Information+Processing+Systems%2C+volume+30%2E+Curran+Associates%2C+Inc%2E)
32. Rohit Sridhar and Diyi Yang. 2022. Explaining toxic text via knowledge enhanced text generation. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 811–826, Seattle, United States. Association for C… [find](https://scholar.google.com/scholar?q=Rohit+Sridhar+and+Diyi+Yang%2E+2022%2E+Explaining+toxic+text+via+knowledge+enhanced+text+generation%2E+In+Proceedings+of+the+2022+Conference+of+the+North+American+Chapter+of+the+Association+for+Computationa)
33. Peng Tang, Pengkai Zhu, Tian Li, Srikar Appalaraju, Vijay Mahadevan, and R. Manmatha. 2024. DEED: Dynamic early exit on decoder for accelerating encoder-decoder transformer models. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 116–131, Mexico City, Mexico. Associati… [find](https://scholar.google.com/scholar?q=Peng+Tang%2C+Pengkai+Zhu%2C+Tian+Li%2C+Srikar+Appalaraju%2C+Vijay+Mahadevan%2C+and+R%2E+Manmatha%2E+2024%2E+DEED%3A+Dynamic+early+exit+on+decoder+for+accelerating+encoder%2Ddecoder+transformer+models%2E+In+Findings+of+the+)
34. Manuel Tonneau, Diyi Liu, Niyati Malhotra, Scott A. Hale, Samuel Fraiberger, Victor Orozco-Olvera, and Paul Röttger. 2025. HateDay: Insights from a global hate speech dataset representative of a day on Twitter. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistic… [find](https://scholar.google.com/scholar?q=Manuel+Tonneau%2C+Diyi+Liu%2C+Niyati+Malhotra%2C+Scott+A%2E+Hale%2C+Samuel+Fraiberger%2C+Victor+Orozco%2DOlvera%2C+and+Paul+R%C3%B6ttger%2E+2025%2E+HateDay%3A+Insights+from+a+global+hate+speech+dataset+representative+of+a+day+o)
35. Marcos Treviso, Ji-Ung Lee, Tianchu Ji, Betty van Aken, Qingqing Cao, Manuel R. Ciosici, Michael Hassid, Kenneth Heafield, Sara Hooker, Colin Raffel, Pedro H. Martins, André F. T. Martins, Jessica Zosa Forde, Peter Milder, Edwin Simpson, Noam Slonim, Jesse Dodge, Emma Strubell, Niranjan Balasubraman… [find](https://scholar.google.com/scholar?q=Marcos+Treviso%2C+Ji%2DUng+Lee%2C+Tianchu+Ji%2C+Betty+van+Aken%2C+Qingqing+Cao%2C+Manuel+R%2E+Ciosici%2C+Michael+Hassid%2C+Kenneth+Heafield%2C+Sara+Hooker%2C+Colin+Raffel%2C+Pedro+H%2E+Martins%2C+Andr%C3%A9+F%2E+T%2E+Martins%2C+Jessica+Zos)
36. Bertie Vidgen and Leon Derczynski. 2020. Directions in abusive language training data, a systematic review: Garbage in, garbage out. PLOS ONE, 15(12):e0243300. [find](https://scholar.google.com/scholar?q=Bertie+Vidgen+and+Leon+Derczynski%2E+2020%2E+Directions+in+abusive+language+training+data%2C+a+systematic+review%3A+Garbage+in%2C+garbage+out%2E+PLOS+ONE%2C+15%2812%29%3Ae0243300%2E)
37. Zeerak Waseem and Dirk Hovy. 2016. Hateful symbols or hateful people? predictive features for hate speech detection on Twitter. In Proceedings of the NAACL Student Research Workshop, pages 88–93, San Diego, California. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Zeerak+Waseem+and+Dirk+Hovy%2E+2016%2E+Hateful+symbols+or+hateful+people%3F+predictive+features+for+hate+speech+detection+on+Twitter%2E+In+Proceedings+of+the+NAACL+Student+Research+Workshop%2C+pages+88%E2%80%9393%2C+San+) [our card](https://md.research.elamur.ai/free/en/papers/waseem-2016-hateful-symbols-or-hateful-people/)
38. Michael Wiegand, Josef Ruppenhofer, and Thomas Kleinbauer. 2019. Detection of Abusive Language: the Problem of Biased Datasets. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short… [find](https://scholar.google.com/scholar?q=Michael+Wiegand%2C+Josef+Ruppenhofer%2C+and+Thomas+Kleinbauer%2E+2019%2E+Detection+of+Abusive+Language%3A+the+Problem+of+Biased+Datasets%2E+In+Proceedings+of+the+2019+Conference+of+the+North+American+Chapter+of+t) [our card](https://md.research.elamur.ai/free/en/papers/wiegand-2019-detection-abusive-language-problem-biased/)
39. Sean Xie, Soroush Vosoughi, and Saeed Hassanpour. 2023. Proto-lm: A prototypical network-based framework for built-in interpretability in large language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 3964–3979, Singapore. Association for Computational Linguis… [find](https://scholar.google.com/scholar?q=Sean+Xie%2C+Soroush+Vosoughi%2C+and+Saeed+Hassanpour%2E+2023%2E+Proto%2Dlm%3A+A+prototypical+network%2Dbased+framework+for+built%2Din+interpretability+in+large+language+models%2E+In+Findings+of+the+Association+for+Comp)
40. Ji Xin, Raphael Tang, Jaejun Lee, Yaoliang Yu, and Jimmy Lin. 2020. DeeBERT: Dynamic early exiting for accelerating BERT inference. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2246–2251, Online. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ji+Xin%2C+Raphael+Tang%2C+Jaejun+Lee%2C+Yaoliang+Yu%2C+and+Jimmy+Lin%2E+2020%2E+DeeBERT%3A+Dynamic+early+exiting+for+accelerating+BERT+inference%2E+In+Proceedings+of+the+58th+Annual+Meeting+of+the+Association+for+Com) [our card](https://md.research.elamur.ai/free/en/papers/xin-2020-deebert-dynamic-early-exiting-accelerating/)
41. Zachary Yang, Nicolas Grenon-Godbout, and Reihaneh Rabbany. 2023. Towards detecting contextual real-time toxicity for in-game chat. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9894–9906, Singapore. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Zachary+Yang%2C+Nicolas+Grenon%2DGodbout%2C+and+Reihaneh+Rabbany%2E+2023%2E+Towards+detecting+contextual+real%2Dtime+toxicity+for+in%2Dgame+chat%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+EMNLP+)
42. Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, Todor Mihaylov, Myle Ott, Sam Shleifer, Kurt Shuster, Daniel Simig, Punit Singh Koura, Anjali Sridhar, Tianlu Wang, and Luke Zettlemoyer. 2022. Opt: Open pre-trai… [find](https://scholar.google.com/scholar?q=Susan+Zhang%2C+Stephen+Roller%2C+Naman+Goyal%2C+Mikel+Artetxe%2C+Moya+Chen%2C+Shuohui+Chen%2C+Christopher+Dewan%2C+Mona+Diab%2C+Xian+Li%2C+Xi+Victoria+Lin%2C+Todor+Mihaylov%2C+Myle+Ott%2C+Sam+Shleifer%2C+Kurt+Shuster%2C+Daniel+S)
43. Yachao Zhao, Bo Wang, Yan Wang, Dongming Zhao, Ruifang He, and Yuexian Hou. 2025. Explicit vs. implicit: Investigating social bias in large language models through self-reflection. In Findings of the Association for Computational Linguistics: ACL 2025, pages 1–12, Vienna, Austria. Association for Co… [find](https://scholar.google.com/scholar?q=Yachao+Zhao%2C+Bo+Wang%2C+Yan+Wang%2C+Dongming+Zhao%2C+Ruifang+He%2C+and+Yuexian+Hou%2E+2025%2E+Explicit+vs%2E+implicit%3A+Investigating+social+bias+in+large+language+models+through+self%2Dreflection%2E+In+Findings+of+the+)
44. Wangchunshu Zhou, Canwen Xu, Tao Ge, Julian McAuley, Ke Xu, and Furu Wei. 2020. Bert loses patience: Fast and robust inference with early exit. In Advances in Neural Information Processing Systems, volume 33, pages 18330–18341. Curran Associates, Inc. 9. [find](https://scholar.google.com/scholar?q=Wangchunshu+Zhou%2C+Canwen+Xu%2C+Tao+Ge%2C+Julian+McAuley%2C+Ke+Xu%2C+and+Furu+Wei%2E+2020%2E+Bert+loses+patience%3A+Fast+and+robust+inference+with+early+exit%2E+In+Advances+in+Neural+Information+Processing+Systems%2C+vo)
45. Language Resource References ElSherief, Mai and Ziems, Caleb and Muchlinski, David and Anupindi, Vaishnavi and Seybolt, Jordyn and De Choudhury, Munmun and Yang, Diyi. 2021. Latent Hatred: A Benchmark for Understanding Implicit Hate Speech. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Language+Resource+References+ElSherief%2C+Mai+and+Ziems%2C+Caleb+and+Muchlinski%2C+David+and+Anupindi%2C+Vaishnavi+and+Seybolt%2C+Jordyn+and+De+Choudhury%2C+Munmun+and+Yang%2C+Diyi%2E+2021%2E+Latent+Hatred%3A+A+Benchmark) [our card](https://md.research.elamur.ai/free/en/papers/elsherief-2021-latent-hatred-benchmark-understanding-implicit/)
46. Mathew, Binny and Saha, Punyajoy and Yimam, Seid Muhie and Biemann, Chris and Goyal, Pawan and Mukherjee, Animesh. 2021. HateXplain: A Benchmark Dataset for Explainable Hate Speech Detection. [find](https://scholar.google.com/scholar?q=Mathew%2C+Binny+and+Saha%2C+Punyajoy+and+Yimam%2C+Seid+Muhie+and+Biemann%2C+Chris+and+Goyal%2C+Pawan+and+Mukherjee%2C+Animesh%2E+2021%2E+HateXplain%3A+A+Benchmark+Dataset+for+Explainable+Hate+Speech+Detection%2E)
47. Sap, Maarten and Gabriel, Saadia and Qin, Lianhui and Jurafsky, Dan and Smith, Noah A. and Choi, Yejin. 2020. Social Bias Frames: Reasoning about Social and Power Implications of Language. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Sap%2C+Maarten+and+Gabriel%2C+Saadia+and+Qin%2C+Lianhui+and+Jurafsky%2C+Dan+and+Smith%2C+Noah+A%2E+and+Choi%2C+Yejin%2E+2020%2E+Social+Bias+Frames%3A+Reasoning+about+Social+and+Power+Implications+of+Language%2E+Association) [our card](https://md.research.elamur.ai/free/en/papers/sap-2020-social-bias-frames-reasoning-about/)
48. Zampieri, Marcos and Malmasi, Shervin and Nakov, Preslav and Rosenthal, Sara and Farra, Noura and Kumar, Ritesh. 2019. Predicting the Type and Target of Offensive Posts in Social Media. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Zampieri%2C+Marcos+and+Malmasi%2C+Shervin+and+Nakov%2C+Preslav+and+Rosenthal%2C+Sara+and+Farra%2C+Noura+and+Kumar%2C+Ritesh%2E+2019%2E+Predicting+the+Type+and+Target+of+Offensive+Posts+in+Social+Media%2E+Association+fo) [our card](https://md.research.elamur.ai/free/en/papers/zampieri-2019-predicting-type-target-offensive-posts/)