Ph.D. Candidate in Computer Science, Shanghai Jiao Tong University
Expected graduation: June 2027. Open to 2027 full-time and research internship opportunities in LLM post-training, alignment, and efficient inference.
I work on large language model training and inference, with a focus on safe distillation, inference-time alignment, efficient decoding, and search-based generation. I conduct part of my Ph.D. research through a joint training arrangement at Shanghai Artificial Intelligence Laboratory.
My research often starts from a question about model behavior and continues through method design, systems implementation, and large-model evaluation.
We explain latent-trait transfer under SFT distillation and reduce malicious-response transfer from 29.55% to 6.45% with probe-space corridor regularization.
DCD removes the amateur model from the serial proposal path while preserving the original contrastive-decoding distribution, reaching 1.65–1.95× average speedups in 8B settings and about 2× at 70B.
IVG combines token-level implicit values with chunk-level search to align frozen target LLMs purely at inference time.
We use small-model preference differences as dense rewards and values for chunk-level search over stronger frozen LLMs, including cross-tokenizer and black-box settings.
A Mixture-of-Experts processor combines neural prediction with program generation and symbolic execution for numerical question answering and natural language inference.
Selected publications and ongoing work, including both lead-author and collaborative research. My name is shown in bold.
NeurIPS 2026 Poster · First author
@article{liu2026subliminal,
title = {Subliminal Learning as Trait-Direction Drift: A Mechanism and Targeted Control under {SFT} Distillation},
author = {Zhixuan Liu and Zhichen Dong and Yuyu Fan and Xiangtian Li and Chao Yang},
journal = {arXiv preprint arXiv:2609.01091},
note = {Accepted to NeurIPS 2026 Main Track as Poster},
year = {2026},
doi = {10.48550/arXiv.2609.01091},
url = {https://arxiv.org/abs/2609.01091}
}EMNLP 2026 Main Conference · First author
@article{liu2026decoupled,
title = {Decoupled Contrastive Decoding via Expert-Aligned Drafting},
author = {Zhixuan Liu and Zhichen Dong and Yuanfu Wang and Chao Yang},
journal = {arXiv preprint arXiv:2608.12913},
note = {Accepted to EMNLP 2026 Main Conference},
year = {2026},
doi = {10.48550/arXiv.2608.12913},
url = {https://arxiv.org/abs/2608.12913}
}Ongoing work · Second author
ICLR 2026 Poster · Second author
@inproceedings{ICLR2026_ccf6d8b4,
title = {Native Reasoning Models: Training Language Models to Reason on Unverifiable Data},
author = {Yuanfu Wang and Zhixuan Liu and Xiangtian Li and Chaochao Lu and Chao Yang},
booktitle = {International Conference on Learning Representations},
year = {2026},
pages = {126317--126343},
url = {https://proceedings.iclr.cc/paper_files/paper/2026/hash/ccf6d8b4a1fe9d9c8192f00c713872ea-Abstract-Conference.html}
}ACL 2026 Findings · Second author
@inproceedings{zhang-etal-2026-adaptive,
title = {Adaptive Prompt Optimization for Open-Ended Tasks: Uncertainty Preference as a Secondary Signal},
author = {Zhang, Shuyang and Liu, Zhixuan and Dong, Zhichen and Zhang, Hao and Lu, Chaochao and Yang, Chao},
booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
year = {2026},
pages = {33883--33891},
doi = {10.18653/v1/2026.findings-acl.1692},
url = {https://aclanthology.org/2026.findings-acl.1692/}
}arXiv 2025 · Contributed the inference-time alignment module in Section 4.1
@article{safework-r1-2025,
title = {SafeWork-R1: Coevolving Safety and Intelligence under the {AI}-45$^{\circ}$ Law},
author = {Yicheng Bao and Guanxu Chen and Mingkang Chen and Yunhao Chen and Chiyu Chen and Lingjie Chen and Sirui Chen and Xinquan Chen and Jie Cheng and Yu Cheng and Dengke Deng and Yizhuo Ding and Dan Ding and Xiaoshan Ding and Yi Ding and Zhichen Dong and Lingxiao Du and Yuyu Fan and Xinshun Feng and Yanwei Fu and Yuxuan Gao and Ruijun Ge and Tianle Gu and Lujun Gui and Jiaxuan Guo and Qianxi He and Yuenan Hou and Xuhao Hu and Hong Huang and Kaichen Huang and Shiyang Huang and Yuxian Jiang and Shanzhe Lei and Jie Li and Lijun Li and Hao Li and Juncheng Li and Xiangtian Li and Yafu Li and Lingyu Li and Xueyan Li and Haotian Liang and Dongrui Liu and Qihua Liu and Zhixuan Liu and Bangwei Liu and Huacan Liu and Yuexiao Liu and Zongkai Liu and Chaochao Lu and Yudong Lu and Xiaoya Lu and Zhenghao Lu and Qitan Lv and Caoyuan Ma and Jiachen Ma and Xiaoya Ma and Zhongtian Ma and Lingyu Meng and Ziqi Miao and Yazhe Niu and Yuezhang Peng and Yuan Pu and Han Qi and Chen Qian and Xingge Qiao and Jingjing Qu and Jiashu Qu and Wanying Qu and Wenwen Qu and Xiaoye Qu and Qihan Ren and Qingnan Ren and Qingyu Ren and Jing Shao and Wenqi Shao and Shuai Shao and Dongxing Shi and Xin Song and Xinhao Song and Yan Teng and Xuan Tong and Yingchun Wang and Xuhong Wang and Shujie Wang and Xin Wang and Yige Wang and Yixu Wang and Yuanfu Wang and Futing Wang and Ruofan Wang and Wenjie Wang and Yajie Wang and Muhao Wei and Xiaoyu Wen and Fenghua Weng and Yuqi Wu and Yingtong Xiong and Xingcheng Xu and Chao Yang and Yue Yang and Yang Yao and Yulei Ye and Zhenyun Yin and Yi Yu and Bo Zhang and Qiaosheng Zhang and Jinxuan Zhang and Yexin Zhang and Yinqiang Zheng and Hefeng Zhou and Zhanhui Zhou and Pengyu Zhu and Qingzi Zhu and Yubo Zhu and Bowen Zhou},
journal = {arXiv preprint arXiv:2507.18576},
year = {2025},
url = {https://arxiv.org/abs/2507.18576}
}ICML 2025 Poster · Third author
@inproceedings{pmlr-v267-dong25p,
title = {Emergent Response Planning in {LLM}s},
author = {Dong, Zhichen and Zhou, Zhanhui and Liu, Zhixuan and Yang, Chao and Lu, Chaochao},
booktitle = {Proceedings of the 42nd International Conference on Machine Learning},
pages = {14301--14320},
year = {2025},
volume = {267},
series = {Proceedings of Machine Learning Research},
publisher = {PMLR},
url = {https://proceedings.mlr.press/v267/dong25p.html}
}NeurIPS 2024 · Co-first author, second-listed
@inproceedings{NEURIPS2024_088d9976,
title = {Weak-to-Strong Search: Align Large Language Models via Searching over Small Language Models},
author = {Zhanhui Zhou and Zhixuan Liu and Jie Liu and Zhichen Dong and Chao Yang and Yu Qiao},
booktitle = {Advances in Neural Information Processing Systems},
volume = {37},
pages = {4819--4851},
year = {2024},
doi = {10.52202/079017-0156},
url = {https://proceedings.neurips.cc/paper_files/paper/2024/hash/088d99765bc121c6df215da7d45bc4e9-Abstract-Conference.html}
}EMNLP 2024 Findings · Co-first author, first-listed
@inproceedings{liu-etal-2024-inference,
title = {Inference-Time Language Model Alignment via Integrated Value Guidance},
author = {Liu, Zhixuan and Zhou, Zhanhui and Wang, Yuanfu and Yang, Chao and Qiao, Yu},
booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2024},
year = {2024},
pages = {4181--4195},
doi = {10.18653/v1/2024.findings-emnlp.242},
url = {https://aclanthology.org/2024.findings-emnlp.242/}
}EMNLP 2022 Findings · First author
@inproceedings{liu-etal-2022-neural,
title = {A Neural-Symbolic Approach to Natural Language Understanding},
author = {Liu, Zhixuan and Wang, Zihao and Lin, Yuan and Li, Hang},
booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2022},
year = {2022},
pages = {2159--2172},
doi = {10.18653/v1/2022.findings-emnlp.158},
url = {https://aclanthology.org/2022.findings-emnlp.158/}
}