Elamur Research⌕
☰
home / papers / 2026 / TeleAI at SemEval-2026 Task 13: Data-Centric Full-Parameter Fine-Tuning with Multi-Level Ensembling for Generated Code Detection

TeleAI at SemEval-2026 Task 13: Data-Centric Full-Parameter Fine-Tuning with Multi-Level Ensembling for Generated Code Detection

Paper · 2026 · Shiquan Wang, Fang Yu, Shuangyong Song et al.

Summary

The paper describes TeleAI's entry to SemEval-2026 Task 13, which focuses on telling human-written source code apart from LLM-generated code. The task has three subtasks:

  • A: binary detection under out-of-distribution shifts.
  • B: attributing code to the specific generator model.
  • C: four-way classification that includes hybrid cases (human code edited by an LLM) and adversarial cases (generation designed to evade detection).

The team's pipeline has four stages:

  1. Data analysis: they examined label and length distributions, duplicated minority-class samples to counter imbalance, and applied uniform truncation and padding.
  2. Prompt selection: they chose per-subtask prompts through iterative validation, pruning weak candidates.
  3. Fine-tuning: they fully fine-tuned Qwen3-30B-A3B-Instruct with several hyperparameter configurations, saving multiple checkpoints.
  4. Ensembling: they merged checkpoints with soft, hard and logits-based voting, plus LightGBM stacking.

Full fine-tuning gave by far the largest improvement over the zero-shot model, and ensembling added smaller but consistent gains. The system ranked first in Subtasks A and B and second in Subtask C. The work shows that careful data handling combined with diverse checkpoint ensembles makes a large LLM a strong code-provenance detector.

Key points

  • Official ranking (test-B, Macro-F1): Subtask A 0.99708 (1st), Subtask B 0.50819 (1st), Subtask C 0.73332 (2nd).
  • Backbone: Qwen3-30B-A3B-Instruct with full-parameter fine-tuning, chosen over parameter-efficient methods. Each subtask was trained separately.
  • Ablation on test-A (zero-shot → fine-tuned → ensemble):
    • Subtask A: 0.4185 → 0.9476 → 0.9956
    • Subtask B: 0.0859 → 0.5146 → 0.5236
    • Subtask C: 0.1475 → 0.8959 → 0.9100
  • Data balancing: imbalance was handled only by repeating minority-class samples, with no synthetic or external data.
  • Ensemble methods: soft voting, hard voting, logits-based fusion, and a second-level LightGBM stacker trained on validation-set probabilities and logits.

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

🏢China TelecomBeijing · China9 works · 16 citations🏢China Telecom (China)Changsha · China8 works · 15 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksInformation Systems · 1,730 works

Cite

Cite as Shiquan Wang, Fang Yu, Shuangyong Song, Yongxiang Li and Xuelong Li. 2026. TeleAI at SemEval-2026 Task 13: Data-Centric Full-Parameter Fine-Tuning with Multi-Level Ensembling for Generated Code Detection. Proceedings of the 20th International Workshop on Semantic Evaluation (2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.semeval-1.38

@inproceedings{wang2026teleai,
  title = {{TeleAI at SemEval-2026 Task 13: Data-Centric Full-Parameter Fine-Tuning with Multi-Level Ensembling for Generated Code Detection}},
  author = {Shiquan Wang, Fang Yu, Shuangyong Song, Yongxiang Li and Xuelong Li},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.38},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.38}
}

References

37 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ankit Aich, Souvik Bhattacharya, and Natalie Parde. 2022. Demystifying neural fake news via linguistic feature-based interpretation. In Proceedings of the 29th International Conference on Computational Linguistics, pages 6586–6599. find
  2. Yuki Arase and Ming Zhou. 2013. Machine translation detection from monolingual web-text. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1597–1607. find
  3. Amrita Bhattacharjee and Huan Liu. 2023. Fighting fire with fire: Can chatgpt detect ai-generated text? arXiv preprint arXiv:2308.01284. — [arXiv:2308.01284](https://arxiv.org/abs/2308.01284) find
  4. Liam Dugan, Daphne Ippolito, Arun Kirubarajan, Sherry Shi, and Chris Callison-Burch. 2023. Real or fake text?: Investigating human ability to detect boundaries between human-written and machine-generated text. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 12763–1… find
  5. Tiziano Fagni, Fabrizio Falchi, Margherita Gambini, Antonio Martella, and Maurizio Tesconi. 2021. Tweepfake: About detecting deepfake tweets. Plos one, 16(5):e0251415.
  6. Chenxi Gu, Chengsong Huang, Xiaoqing Zheng, Kai-Wei Chang, and Cho-Jui Hsieh. 2022. Watermarking pre-trained language models with backdooring. arXiv preprint arXiv:2210.07543. — [arXiv:2210.07543](https://arxiv.org/abs/2210.07543) find
  7. Ahmed Abdeen Hamed and Xindong Wu. 2023. Improving detection of chatgpt-generated fake science using real publication text: Introducing xfakebibs a supervised-learning network algorithm. find
  8. Zhongjiang He, Zihan Wang, Xinzhang Liu, Shixuan Liu, Yitong Yao, Yuyao Huang, Xuelong Li, Yongxiang Li, Zhonghao Che, Zhaoxi Zhang, and 1 others. 2024. Telechat technical report. arXiv preprint arXiv:2401.03804.
  9. Daphne Ippolito, Daniel Duckworth, Chris Callison-Burch, and Douglas Eck. 2020. Automatic detection of generated text is easiest when humans are fooled. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1808– 1822.
  10. John Kirchenbauer, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers, and Tom Goldstein. 2023. A watermark for large language models. In International Conference on Machine Learning, pages 17061–17084. PMLR.
  11. Xiang Li, Yiqun Yao, Xin Jiang, Xuezhi Fang, Chao Wang, Xinzhang Liu, Zihan Wang, Yu Zhao, Xin Wang, Yuyao Huang, and 1 others. 2024a. 52b to 1t: Lessons learned via tele-flm series. arXiv preprint arXiv:2407.02783. — [arXiv:2407.02783](https://arxiv.org/abs/2407.02783) find
  12. Xiang Li, Yiqun Yao, Xin Jiang, Xuezhi Fang, Chao Wang, Xinzhang Liu, Zihan Wang, Yu Zhao, Xin Wang, Yuyao Huang, and 1 others. 2024b. Tele-flm technical report. arXiv preprint arXiv:2404.16645. — [arXiv:2404.16645](https://arxiv.org/abs/2404.16645) find
  13. Zhongqiu Li, Shiquan Wang, Ruiyu Fang, Mengjiao Bao, Zhenhe Wu, Shuangyong Song, Yongxiang Li, and Zhongjiang He. 2025. Mr-uie: multi-perspective reasoning with reinforcement learning for universal information extraction. Vicinagearth, 2(1):17. find
  14. Aiwei Liu, Leyi Pan, Xuming Hu, Shiao Meng, and Lijie Wen. A semantic invariant robust watermark for large language models, 2024. URL https://arxiv.org/abs/2310.06356. — [arXiv:2310.06356](https://arxiv.org/abs/2310.06356) ↗
  15. Xinzhang Liu, Chao Wang, Zhihao Yang, Zhuo Jiang, Xuncheng Zhao, Haoran Wang, Lei Li, Dongdong He, Luobin Liu, Kaizhe Yuan, and 1 others. 2025. Training report of telechat3-moe. arXiv preprint arXiv:2512.24157. — [arXiv:2512.24157](https://arxiv.org/abs/2512.24157) find
  16. Yaojie Lu, Qing Liu, Dai Dai, Xinyan Xiao, Hongyu Lin, Xianpei Han, Le Sun, and Hua Wu. 2022. Unified structure generation for universal information extraction. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5755–5772. find
  17. Daniil Orel, Dilshod Azizov, Indraneil Paul, Yuxia Wang, Iryna Gurevych, and Preslav Nakov. 2026. SemEval-2026 task 13: Detecting machine-generated code with multiple programming languages, generators, and application scenarios. In Proceedings of the 20th International Workshop on Semantic Evaluatio…
  18. Juan Diego Rodriguez, Todd Hay, David Gros, Zain Shamsi, and Ravi Srinivasan. 2022. Cross-domain detection of gpt-2-generated technical text. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1213–… find
  19. Lingling Shi, Haoyu Jin, Ruiyu Fang, Shuangyong Song, Jinsong Su, Yongxiang Li, and Xuelong Li. 2026. Cemtcontrollable element-oriented machine translation via structured linguistic reasoning. In Findings of the association for computational linguistics ACL 2026. find
  20. Shuangyong Song, Haiqing Chen, and Zhiwei Shi. 2017. Intention classification of user queries in intelligent customer service system. In 2017 International Conference on Asian Language Processing (IALP), pages 83–86. IEEE. find
  21. Shuangyong Song and Yao Meng. 2015. Classifying and ranking microblogging hashtags with news categories. In 2015 IEEE 9th International Conference on Research Challenges in Information Science (RCIS), pages 540–541. IEEE. find
  22. Shuangyong Song, Chao Wang, Haiqing Chen, and Huan Chen. 2020a. Tcnn: Triple convolutional neural network models for retrieval-based question answering system in e-commerce. In Companion Proceedings of the Web Conference 2020, pages 844– 845. find
  23. Shuangyong Song, Chao Wang, Siyang Liu, Haiqing Chen, Huan Chen, and Hongyun Bao. 2020b. Sentiment analysis technologies in alime—an intelligent assistant for e-commerce. International Journal of Asian Language Processing, 30(04):2050016. find
  24. Jinyan Su, Terry Zhuo, Di Wang, and Preslav Nakov. 2023. Detectllm: Leveraging log rank information for zero-shot detection of machine-generated text. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 12395–12412.
  25. Shiquan Wang, Ruiyu Fang, Mengxiang Li, Zhongjiang He, and Shuangyong Song. 2025a. When less is more: Minimal prompts with lora for llm text detection. In CCF International Conference on Natural Language Processing and Chinese Computing, pages 275–283. Springer. find
  26. Zihan Wang, Xinzhang Liu, Yitong Yao, Chao Wang, Yu Zhao, Zhihao Yang, Wenmin Deng, Kaipeng Jia, Jiaxin Peng, Yuyao Huang, and 1 others. 2025b. Technical report of telechat2, telechat2. 5 and t1. arXiv preprint arXiv:2507.18013. — [arXiv:2507.18013](https://arxiv.org/abs/2507.18013) find
  27. Zihan Wang, Yitong Yao, Li Mengxiang, Zhongjiang He, Chao Wang, Shuangyong Song, and 1 others. 2024. Telechat: An open-source billingual large language model. In Proceedings of the 10th SIGHAN Workshop on Chinese Language Processing (SIGHAN-10), pages 10–20. find
  28. Luoxuan Weng, Minfeng Zhu, Kam Kwai Wong, Shi Liu, Jiashun Sun, Hang Zhu, Dongming Han, and Wei Chen. 2023. Towards an understanding and explanation for mixed-initiative artificial scientific text detection. arXiv preprint arXiv:2304.05011. others. 2025b. Tablereasoner: Advancing table reasoning fra… — [arXiv:2304.05011](https://arxiv.org/abs/2304.05011) find
  29. Sishi Xiong, Yu Zhao, Jie Zhang, Li Mengxiang, Zhongjiang He, Xuelong Li, and Shuangyong Song. 2024. Dual prompt tuning based contrastive learning for hierarchical text classification. In Findings of the association for computational linguistics ACL 2024, pages 12146–12158. find
  30. Jian Yang, Xianglong Liu, Weifeng Lv, Ken Deng, Shawn Guo, Lin Jing, Yizhi Li, Shark Liu, Xianzhen Luo, Yuyu Luo, and 1 others. 2025. From code foundation models to agents and applications: A comprehensive survey and practical guide to code intelligence. arXiv preprint arXiv:2511.18538. — [arXiv:2511.18538](https://arxiv.org/abs/2511.18538) find
  31. KiYoon Yoo, Wonhyuk Ahn, Jiho Jang, and Nojun Kwak. 2023. Robust multi-bit natural language watermarking through invariant features. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2092–2115. find
  32. Xiao Yu, Yuang Qi, Kejiang Chen, Guoqiang Chen, Xi Yang, Pengyuan Zhu, Weiming Zhang, and Nenghai Yu. 2023. Gpt paternity test: Gpt generated text detection with gpt genetic inheritance. CoRR. find
  33. Deji Zhao, Donghong Han, Jia Wu, Zhongjiang He, Bo Ning, Ye Yuan, Yongxiang Li, Chao Wang, and Shuangyong Song. 2025. Enhancing math reasoning ability of large language models via computation logic graphs. Knowledge-Based Systems, 325:113905. find
  34. Kangxi Wu, Liang Pang, Huawei Shen, Xueqi Cheng, and Tat-Seng Chua. 2023. Llmdet: A third party large language models generated text detection tool. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 2113–2133. find
  35. Hongrui Xing, Xinzhang Liu, Zhuo Jiang, Zhihao Yang, Yitong Yao, Zihan Wang, Wenmin Deng, Chao Wang, Shuangyong Song, Wang Yang, and 1 others. 2025. Llmsr@ xllm25: A language model-based pipeline for structured reasoning data construction. In Proceedings of the 1st Joint Workshop on Large Language M… find
  36. Sishi Xiong, Mengxiang Li, Dakai Wang, Yu Zhao, Jie Zhang, Changzai Pan, Haowei He, Xiangyu Li, Wenhan Chang, Zhongjiang He, and 1 others. 2025a. Teleai at semeval-2025 task 8: Advancing table reasoning framework with large language models. In Proceedings of the 19th International Workshop on Semant… find
  37. Sishi Xiong, Dakai Wang, Yu Zhao, Jie Zhang, Changzai Pan, Haowei He, Xiangyu Li, Wenhan Chang, Zhongjiang He, Shuangyong Song, and 1 find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-teleai-semeval-2026-task-13/