Elamur Research⌕
☰
home / papers / 2026 / wangkongqiang@EEUCA 2026: Understanding Toxic Behavioral Intent in Gaming Chat Logs

wangkongqiang@EEUCA 2026: Understanding Toxic Behavioral Intent in Gaming Chat Logs

Paper · 2026 · Kongqiang Wang, Peng Zhang and Quingli Tan

Summary

This system paper describes a team's entry in the EEUCA 2026 shared task on detecting toxic intent in multiplayer game chat. Every player message is classified into one of six categories: a benign Non-toxic class or one of five toxic classes, namely Threats, Extremism, Hate and Harassment, Insults and Flaming, and Other Offensive Texts. The data come from the GameTox dataset (NAACL 2025), which contains World of Tanks chat. Of 53,000 collected utterances, 42,963 English ones carry intent labels, and the label distribution is heavily skewed toward non-toxic messages.

The authors compared two approaches:

  • Encoder models: fine-tuned classifiers (ALBERT, BERT, ERNIE 2.0, RoBERTa variants), some with RNN, CNN or LSTM heads, plus a hard-voting ensemble.
  • Instruction-tuned LLMs: fine-tuned Qwen2 at the 1.5B and 7B sizes.

The Qwen2 7B model clearly outperformed the encoder baselines, reaching a test Macro F1 of 0.5776 and placing 8th. The work shows that LLM fine-tuning helps with rare toxic categories under strong imbalance. Even so, the task remains hard, because the strongest team scored about 0.70 Macro F1.

Key points

  • Dataset: GameTox (World of Tanks chat logs). Non-toxic messages make up about 80.7% of labeled utterances, while Threats has only 53 examples and Extremism only 23.
  • Best submission: fine-tuned Qwen2 7B, with test Macro F1 0.5776, Accuracy 0.9075, Macro Precision 0.6847 and Macro Recall 0.5343. It ranked 8th of 35 teams on the leaderboard.
  • Encoder baselines: these stayed far behind. The best was simcse-roberta-lstm at Macro F1 0.4234, followed by bert-base-uncased at 0.4205. Qwen2 1.5B reached 0.4525.
  • Ensemble: the hard-voting bagging ensemble of four RoBERTa-based models did not improve Macro F1, scoring 0.372.
  • Stated limitations: the system ignores conversational context, and limited compute restricted the authors to the 7B model. The code is available on GitHub.

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

🎓Yunnan UniversityKunming · China20 works · 40 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksSocial Sciences · 1,954 worksCommunication · 104 works

🧭 Topics 1Research subjects and languages the work is about.Bias, fairness & safety · 587 works

Cite

Cite as Kongqiang Wang, Peng Zhang and Quingli Tan. 2026. wangkongqiang@EEUCA 2026: Understanding Toxic Behavioral Intent in Gaming Chat Logs. Proceedings of the 9th Workshop on Event Extraction and Understanding: Challenges and Applications (EEUCA 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eeuca-1.11

@inproceedings{wang2026wangkongqiang,
  title = {{wangkongqiang@EEUCA 2026: Understanding Toxic Behavioral Intent in Gaming Chat Logs}},
  author = {Kongqiang Wang, Peng Zhang and Quingli Tan},
  booktitle = {Proceedings of the 9th Workshop on Event Extraction and Understanding: Challenges and Applications (EEUCA 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.eeuca-1.11},
  url = {https://doi.org/10.18653/v1/2026.eeuca-1.11}
}

References

19 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Aashish Bhandari, Siddhant B Shah, Surendrabikram Thapa, Usman Naseem, and Mehwish Nasim. 2023. Crisishatemm: Multimodal analysis of directed and undirected hate speech in text-embedded images from russia-ukraine conflict. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco… find
  2. Jeremy Blackburn and Haewoon Kwak. 2014. Stfu noob! predicting crowdsourced decisions on toxic behavior in online games. In Proceedings of the 23rd international conference on World wide web, page 877–888. find
  3. Jing Qian, Anna Bethke, Yinyin Liu, Elizabeth Belding, and William Yang Wang. 2019. A benchmark dataset for learning to intervene in online hate speech.
  4. Antigoni Founta, Constantinos Djouvas, Despoina Chatzakou, Ilias Leontiadis, Jeremy Blackburn, Gianluca Stringhini, Athena Vakali, Michael Sirivianos, and Nicolas Kourtellis. 2018. Large scale crowdsourcing and characterization of twitter abusive behavior. Proceedings of the International AAAI Confe… find
  5. Ali Hurriyetoglu, Surendrabikram Thapa, Hristo Tanev, and Surabhi Adhikari. 2025. Findings and insights from the 8th workshop on challenges and applications of automated extraction of socio-political events from text. In Proceedings of the 8th Workshop on Challenges and Applications of Automated Ext… find
  6. Ali Hürriyetoğlu, Surendrabikram Thapa, Hristo Tanev, Laxmi Thapa, and Surabhi Adhikari. 2026. Overview of the workshop on event extraction and understanding: Challenges and applications. In Proceedings of the 9th Workshop on Event Extraction and Understanding: Challenges and Applications (EEUCA). find
  7. Binny Mathew, Punyajoy Saha, Seid Muhie Yimam, Chris Biemann, Pawan Goyal, and Animesh Mukherjee. 2020. Hatexplain: A benchmark dataset for explainable hate speech detection. In AAAI Conference on Artificial Intelligence. find
  8. Marcus Märtens, Siqi Shen, Alexandru Iosup, and Fer nando Kuipers. 2015. Toxicity detection in multiplayer online games. In 2015 International Workshop on Network and Systems Support for Games (NetGames), page 1–6. IEEE. find
  9. Usman Naseem, Shuvam Shiwakoti, Siddhant Bikram Shah, Surendrabikram Thapa, and Qi Zhang. 2025. Gametox: A comprehensive dataset and analysis for enhanced toxicity detection in online gaming communities. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association… find
  10. Mustafa Oz, Akan Yanik, and Mikail Batu. 2023. Under the shadow of culture and politics: Understanding lgbtq social media activists’ perceptions, concerns, and strategies. Social Media + Society, 9(3). find
  11. Anil Singh Parihar, Surendrabikram Thapa, and Sushruti Mishra. 2021. Hate speech detection using natural language processing: Applications and challenges. In 2021 5th International Conference on Trends in Electronics and Informatics (ICOEI), pages 1302– 1308. IEEE. find
  12. Semiu Salawu, Jo Lumsden, and Yulan He. 2021. A large-scale English multi-label Twitter dataset for cyberbullying and online abuse detection. In Proceedings of the 5th Workshop on Online Abuse and Harms (WOAH 2021), pages 146–156, Online. Association for Computational Linguistics.
  13. Wessel Stoop, Florian Kunneman, Antal van den Bosch, and Ben Miller. 2019. Detecting harassment in real-time as conversations develop. In Proceedings of the Third Workshop on Abusive Language Online, pages 19–24. find
  14. Surendrabikram Thapa, Farhan Ahmad Jafri, Ali Hürriyetoğlu, Francielle Vargas, Roy Ka Wei Lee, and Usman Naseem. 2023. Multimodal hate speech event detection-shared task 4. In CASE 2023-Proceedings of the 6th Workshop on Challenges and Applications of Automated Extraction of Socio-Political Events… find
  15. Surendrabikram Thapa, Kritesh Rauniyar, Farhan Ahmad Jafri, Hariram Veeramani, Raghav Jain, Sandesh Jain, Francielle Vargas, Ali Hürriyetoğlu, and Usman Naseem. 2024. Extended multimodal hate speech event detection during russia-ukraine crisis-shared task at case 2024. In 7th Workshop on Challenge… find
  16. Surendrabikram Thapa, Shuvam Shiwakoti, Siddhant Bikram Shah, Surabhi Adhikari, Hariram Veeramani, Mehwish Nasim, and Usman Naseem. 2025. Large language models (llm) in computational social science: prospects, current state, and challenges. Social Network Analysis and Mining, 15(1):1–30. find
  17. Surendrabikram Thapa, Shuvam Shiwakoti, Siddhant Bikram Shah, Kritesh Rauniyar, Laxmi Thapa, Surabhi Adhikari, Kristina T. Johnson, Ali Hürriyetoğlu, Hristo Tanev, and Usman Naseem. 2026. Understanding toxic behavior in gaming communities using ai to promote healthier digital spaces. In Proceeding… find
  18. Thilini Wijesiriwardene, Hale Inan, Ugur Kursuncu, Manas Gaur, Valerie L. Shalin, Krishnaprasad Thirunarayan, Amit Sheth, and I. Budak Arpinar. 2020. Alone: A dataset for toxic behavior among adolescents on twitter. In Social Informatics, pages 427–439, Cham. Springer International Publishing. find
  19. Marcos Zampieri, Shervin Malmasi, Preslav Nakov, Sara Rosenthal, Noura Farra, and Ritesh Kumar. 2019. Predicting the type and target of offensive posts in social media.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-wangkongqiang-eeuca-2026-understanding-toxic/