publications

2026

1. Nat. Mach. Intell.

   Proposing and solving olympiad geometry with guided tree search

   Chi Zhang, Jiajun Song, Siyu Li, and 5 more authors

   Nature Machine Intelligence, 2026

   arXiv
2. HDSR

   Uncovering hidden geometry in Transformers via disentangling position and context

   Jiajun Song, and Yiqiao Zhong

   Harvard Data Science Review, 2026

   arXiv Bib Blog Code

   @article{song2023uncovering,
     title = {Uncovering hidden geometry in Transformers via disentangling position and context},
     author = {Song, Jiajun and Zhong, Yiqiao},
     journal = {Harvard Data Science Review},
     volume = {8},
     number = {2},
     year = {2026},
   }

3. ACL

   MARCH: Multi-Agent Reinforced Check for Hallucination

   Zhuo Li, Yupeng Zhang, Pengyu Cheng, and 8 more authors

   In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics, 2026

   arXiv Bib

   @inproceedings{li2026march,
     title = {MARCH: Multi-Agent Reinforced Check for Hallucination},
     author = {Li, Zhuo and Zhang, Yupeng and Cheng, Pengyu and Song, Jiajun and Zhou, Mengyu and Li, Hao and Hu, Shujie and Qin, Yu and Zhao, Erchao and Jiang, Xiaoxi and Jiang, Guanjun},
     booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics},
     year = {2026},
   }

4. AAAI

   Mind the gap: The divergence between human and llm-generated tasks

   Yi-Long Lu, Jiajun Song, Chunhui Zhang, and 1 more author

   In Proceedings of the AAAI Conference on Artificial Intelligence, 2026

   Bib

   @inproceedings{lu2026mind,
     title = {Mind the gap: The divergence between human and llm-generated tasks},
     author = {Lu, Yi-Long and Song, Jiajun and Zhang, Chunhui and Wang, Wei},
     booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
     volume = {40},
     number = {3},
     pages = {1964--1972},
     year = {2026},
   }

5. Preprint

   Clipo: Contrastive learning in policy optimization generalizes rlvr

   Sijia Cui, Pengyu Cheng, Jiajun Song, and 6 more authors

   arXiv preprint arXiv:2603.10101, 2026

   arXiv Bib

   @article{cui2026clipo,
     title = {Clipo: Contrastive learning in policy optimization generalizes rlvr},
     author = {Cui, Sijia and Cheng, Pengyu and Song, Jiajun and Gai, Yongbo and Zhang, Guojun and Yu, Zhechao and Lin, Jianhe and Jiang, Xiaoxi and Jiang, Guanjun},
     journal = {arXiv preprint arXiv:2603.10101},
     year = {2026},
   }

6. Preprint

   GD^2PO: Mitigating multi-reward conflicts via group-dynamic reward-decoupled policy optimization

   Haotian Liu, Yihao Liu, Jingwei Ni, and 11 more authors

   arXiv preprint arXiv:2606.16771, 2026

   arXiv Bib

   @article{liu2026gdpo,
     title = {GD$^2$PO: Mitigating multi-reward conflicts via group-dynamic reward-decoupled policy optimization},
     author = {Liu, Haotian and Liu, Yihao and Ni, Jingwei and Huang, Siyuan and Liu, Xinpeng and Cheng, Pengyu and Song, Jiajun and Ding, Ruijin and Li, Junfeng and Yu, Zhechao and Zhou, Mengyu and Xu, Hongteng and Jiang, Xiaoxi and Jiang, Guanjun},
     journal = {arXiv preprint arXiv:2606.16771},
     year = {2026},
   }

2025

1. PNAS

   Out-of-distribution generalization via composition: a lens through induction heads in transformers

   Jiajun Song, Zhuoyan Xu, and Yiqiao Zhong

   Proceedings of the National Academy of Sciences, 2025

   arXiv Blog Code
2. Preprint

   Do theory of mind benchmarks need explicit human-like reasoning in language models?

   Yi-Long Lu, Chunhui Zhang, Jiajun Song, and 2 more authors

   arXiv preprint arXiv:2504.01698, 2025

   arXiv Bib

   @article{lu2025theory,
     title = {Do theory of mind benchmarks need explicit human-like reasoning in language models?},
     author = {Lu, Yi-Long and Zhang, Chunhui and Song, Jiajun and Fan, Lifeng and Wang, Wei},
     journal = {arXiv preprint arXiv:2504.01698},
     year = {2025},
   }

3. Preprint

   A unified representation underlying the judgment of large language models

   Yi-Long Lu, Jiajun Song, and Wei Wang

   arXiv preprint arXiv:2510.27328, 2025

   arXiv Bib

   @article{lu2025unified,
     title = {A unified representation underlying the judgment of large language models},
     author = {Lu, Yi-Long and Song, Jiajun and Wang, Wei},
     journal = {arXiv preprint arXiv:2510.27328},
     year = {2025},
   }