Publications

2026

ArXiv

Skill Training with Corruption and Reconstruction Loop

Mo Li, Zixin Yin, Qihao Wu, Ting Cao, Yunxin Liu, and Heung-Yeung Shum

arXiv Preprint, 2026

@misc{li2026skilltraining,
  title = {Skill Training with Corruption and Reconstruction Loop},
  author = {Li, Mo and Yin, Zixin and Wu, Qihao and Cao, Ting and Liu, Yunxin and Shum, Heung-Yeung},
  year = {2026},
  eprint = {2607.27557},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
  url = {https://arxiv.org/abs/2607.27557},
  journal = {arXiv Preprint}
}

NeurIPS

Learning to Commit: Next-Commit Prediction via Online Supervised Contrastive Reflection

Mo Li, L. H. Xu, Qitai Tan, Ting Cao, and Yunxin Liu

In Proceedings of Advances in Neural Information Processing Systems (NeurIPS), 2026

Bib

@inproceedings{li2026learningcommitgeneratingorganic,
  title = {Learning to Commit: Next-Commit Prediction via Online Supervised Contrastive Reflection},
  author = {Li, Mo and Xu, L. H. and Tan, Qitai and Cao, Ting and Liu, Yunxin},
  year = {2026},
  eprint = {2603.26664},
  archiveprefix = {arXiv},
  primaryclass = {cs.SE},
  url = {https://arxiv.org/abs/2603.26664},
  booktitle = {Proceedings of Advances in Neural Information Processing Systems (NeurIPS)}
}

Tech Report

Kimi K2.5: Visual Agentic Intelligence

Kimi Team

arXiv Preprint, 2026

Bib Code

@misc{kimiteam2026kimik25,
  title = {Kimi K2.5: Visual Agentic Intelligence},
  author = {Team, Kimi},
  year = {2026},
  eprint = {2602.02276},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
  url = {https://arxiv.org/abs/2602.02276},
  journal = {arXiv Preprint}
}

ICLR

Sculptor: Empowering LLMs with Cognitive Agency via Active Context Management

Mo Li, L. H. Xu, Qitai Tan, Long Ma, Ting Cao, and Yunxin Liu

In Proceedings of the International Conference on Learning Representations (ICLR), 2026

Bib

@inproceedings{li2025sculptorempoweringllmscognitive,
  title = {Sculptor: Empowering LLMs with Cognitive Agency via Active Context Management},
  author = {Li, Mo and Xu, L. H. and Tan, Qitai and Ma, Long and Cao, Ting and Liu, Yunxin},
  year = {2026},
  eprint = {2508.04664},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
  url = {https://arxiv.org/abs/2508.04664},
  booktitle = {Proceedings of the International Conference on Learning Representations (ICLR)}
}

ICLR

MMSI-Bench: A Benchmark for Multi-Image Spatial Intelligence

Sihan Yang, Runsen Xu, Yiman Xie, Sizhe Yang, Mo Li, Jingli Lin, Chenming Zhu, Xiaochen Chen, Haodong Duan, Xiangyu Yue, Dahua Lin, Tai Wang, and Jiangmiao Pang

In Proceedings of the International Conference on Learning Representations (ICLR), 2026

Bib Code Website

@inproceedings{yang2025mmsibenchbenchmarkmultiimagespatial,
  title = {MMSI-Bench: A Benchmark for Multi-Image Spatial Intelligence},
  author = {Yang, Sihan and Xu, Runsen and Xie, Yiman and Yang, Sizhe and Li, Mo and Lin, Jingli and Zhu, Chenming and Chen, Xiaochen and Duan, Haodong and Yue, Xiangyu and Lin, Dahua and Wang, Tai and Pang, Jiangmiao},
  year = {2026},
  eprint = {2505.23764},
  archiveprefix = {arXiv},
  primaryclass = {cs.CV},
  url = {https://arxiv.org/abs/2505.23764},
  booktitle = {Proceedings of the International Conference on Learning Representations (ICLR)}
}

2025

TMLR

NeedleBench: Evaluating LLM Retrieval and Reasoning Across Varying Information Densities

Mo Li, Songyang Zhang, Taolin Zhang, Haodong Duan, Yunxin Liu, and Kai Chen

Transactions on Machine Learning Research (TMLR), 2025

Bib PDF Website

@article{li2025needlebench,
  title = {NeedleBench: Evaluating {LLM} Retrieval and Reasoning Across Varying Information Densities},
  author = {Li, Mo and Zhang, Songyang and Zhang, Taolin and Duan, Haodong and Liu, Yunxin and Chen, Kai},
  journal = {Transactions on Machine Learning Research (TMLR)},
  issn = {2835-8856},
  year = {2025},
  url = {https://openreview.net/forum?id=cEvmIKsRw0},
}

NeurIPS

SynTSBench: Rethinking Temporal Pattern Learning in Deep Learning Models for Time Series

Qitai Tan^*, Yiyun Chen^*, Mo Li^*, Ruiwen Gu, Yilin Su, and Xiao-Ping Zhang

In Proceedings of Advances in Neural Information Processing Systems (NeurIPS), 2025

Bib Code Website

@inproceedings{tan2025syntsbench,
  title = {Syn{TSB}ench: Rethinking Temporal Pattern Learning in Deep Learning Models for Time Series},
  author = {Tan, Qitai and Chen, Yiyun and Li, Mo and Gu, Ruiwen and Su, Yilin and Zhang, Xiao-Ping},
  booktitle = {Proceedings of Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2025},
  url = {https://openreview.net/forum?id=rNw128KrYU},
}

ACL
Condor: Enhance LLM Alignment with Knowledge-Driven Data Synthesis and Refinement

Maosong Cao, Taolin Zhang, Mo Li, Chuyu Zhang, Yunxin Liu, Conghui He, Haodong Duan, Songyang Zhang, and Kai Chen

In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL), Jul 2025

Abs DOI Bib Code Website

The quality of Supervised Fine-Tuning (SFT) data plays a critical role in enhancing the conversational capabilities of Large Language Models (LLMs). However, the availability of high-quality human-annotated SFT data has become a significant bottleneck for LLMs, necessitating a greater reliance on synthetic training data. In this work, we introduce Condor, a two-stage synthetic data generation framework that incorporates World Knowledge Trees and Self-Reflection Refinement to produce high-quality SFT data at scale. Our experimental results demonstrate that a base model fine-tuned on only 20K Condor-generated samples achieves superior performance compared to instruct model trained with RLHF. The additional refinement stage in Condor further enables iterative self-improvement for LLMs at various scales (up to 72B), validating the effectiveness of our approach. Furthermore, our investigation into the scaling of synthetic data in post-training reveals substantial unexplored potential for performance improvements, opening promising avenues for future research.
@inproceedings{maosongcao-etal-2025-condor, title = {Condor: Enhance {LLM} Alignment with Knowledge-Driven Data Synthesis and Refinement}, author = {Cao, Maosong and Zhang, Taolin and Li, Mo and Zhang, Chuyu and Liu, Yunxin and He, Conghui and Duan, Haodong and Zhang, Songyang and Chen, Kai}, editor = {Che, Wanxiang and Nabende, Joyce and Shutova, Ekaterina and Pilehvar, Mohammad Taher}, booktitle = {Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)}, month = jul, year = {2025}, address = {Vienna, Austria}, publisher = {Association for Computational Linguistics}, url = {https://aclanthology.org/2025.acl-long.1091/}, doi = {10.18653/v1/2025.acl-long.1091}, pages = {22392--22412}, isbn = {979-8-89176-251-0}, }

2024

Tech Report

InternLM2 Technical Report

Zheng Cai, Maosong Cao, Haojiong Chen, Kai Chen, Keyu Chen, Xin Chen, Xun Chen, Zehui Chen, Zhi Chen, Pei Chu, Xiaoyi Dong, Haodong Duan, Qi Fan, Zhaoye Fei, Yang Gao , and 85 more authors

arXiv Preprint, Jul 2024

Bib Code Website

@misc{cai2024internlm2technicalreport,
  title = {InternLM2 Technical Report},
  author = {Cai, Zheng and Cao, Maosong and Chen, Haojiong and Chen, Kai and Chen, Keyu and Chen, Xin and Chen, Xun and Chen, Zehui and Chen, Zhi and Chu, Pei and Dong, Xiaoyi and Duan, Haodong and Fan, Qi and Fei, Zhaoye and Gao, Yang and Ge, Jiaye and Gu, Chenya and Gu, Yuzhe and Gui, Tao and Guo, Aijia and Guo, Qipeng and He, Conghui and Hu, Yingfan and Huang, Ting and Jiang, Tao and Jiao, Penglong and Jin, Zhenjiang and Lei, Zhikai and Li, Jiaxing and Li, Jingwen and Li, Linyang and Li, Shuaibin and Li, Wei and Li, Yining and Liu, Hongwei and Liu, Jiangning and Hong, Jiawei and Liu, Kaiwen and Liu, Kuikun and Liu, Xiaoran and Lv, Chengqi and Lv, Haijun and Lv, Kai and Ma, Li and Ma, Runyuan and Ma, Zerun and Ning, Wenchang and Ouyang, Linke and Qiu, Jiantao and Qu, Yuan and Shang, Fukai and Shao, Yunfan and Song, Demin and Song, Zifan and Sui, Zhihao and Sun, Peng and Sun, Yu and Tang, Huanze and Wang, Bin and Wang, Guoteng and Wang, Jiaqi and Wang, Jiayu and Wang, Rui and Wang, Yudong and Wang, Ziyi and Wei, Xingjian and Weng, Qizhen and Wu, Fan and Xiong, Yingtong and Xu, Chao and Xu, Ruiliang and Yan, Hang and Yan, Yirong and Yang, Xiaogui and Ye, Haochen and Ying, Huaiyuan and Yu, Jia and Yu, Jing and Zang, Yuhang and Zhang, Chuyu and Zhang, Li and Zhang, Pan and Zhang, Peng and Zhang, Ruijie and Zhang, Shuo and Zhang, Songyang and Zhang, Wenjian and Zhang, Wenwei and Zhang, Xingcheng and Zhang, Xinyue and Zhao, Hui and Zhao, Qian and Zhao, Xiaomeng and Zhou, Fengzhe and Zhou, Zaida and Zhuo, Jingming and Zou, Yicheng and Qiu, Xipeng and Qiao, Yu and Lin, Dahua},
  year = {2024},
  eprint = {2403.17297},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
  url = {https://arxiv.org/abs/2403.17297},
  journal = {arXiv Preprint}
}

Tech Report

VLMEvalKit: An Open-Source Toolkit for Evaluating Large Multi-Modality Models

Haodong Duan, Xinyu Fang, Junming Yang, Xiangyu Zhao, Yuxuan Qiao, Mo Li, Amit Agarwal, Zhe Chen, Lin Chen, Yuan Liu, Yubo Ma, Hailong Sun, Yifan Zhang, Shiyin Lu, Tack Hwa Wong , and 17 more authors

In Proceedings of the ACM International Conference on Multimedia (ACM MM), Jul 2024

Bib Code Website

@inproceedings{duan2024vlmevalkitopensourcetoolkitevaluating,
  title = {VLMEvalKit: An Open-Source Toolkit for Evaluating Large Multi-Modality Models},
  author = {Duan, Haodong and Fang, Xinyu and Yang, Junming and Zhao, Xiangyu and Qiao, Yuxuan and Li, Mo and Agarwal, Amit and Chen, Zhe and Chen, Lin and Liu, Yuan and Ma, Yubo and Sun, Hailong and Zhang, Yifan and Lu, Shiyin and Wong, Tack Hwa and Wang, Weiyun and Zhou, Peiheng and Li, Xiaozhe and Fu, Chaoyou and Cui, Junbo and Chen, Jixuan and Song, Enxin and Mao, Song and Ding, Shengyuan and Liang, Tianhao and Zhang, Zicheng and Dong, Xiaoyi and Zang, Yuhang and Zhang, Pan and Wang, Jiaqi and Lin, Dahua and Chen, Kai},
  year = {2024},
  eprint = {2407.11691},
  archiveprefix = {arXiv},
  primaryclass = {cs.CV},
  url = {https://arxiv.org/abs/2407.11691},
  booktitle = {Proceedings of the ACM International Conference on Multimedia (ACM MM)}
}

2023

Tech Report

OpenCompass: A Universal Evaluation Platform for Large Language Models

Maosong Cao, Kai Chen, Haodong Duan, Yixiao Fang, Zhiwei Fei, Tong Gao, Ge Jiaye, Mo Li, Hongwei Liu, Junnan Liu, Yuan Liu, Chengqi Lyu, Han Lyu, Ningsheng Ma, Zerun Ma , and 15 more authors

arXiv Preprint, Jul 2023

Bib Code Website

@misc{cao2023opencompass,
  title = {OpenCompass: A Universal Evaluation Platform for Large Language Models},
  author = {Cao, Maosong and Chen, Kai and Duan, Haodong and Fang, Yixiao and Fei, Zhiwei and Gao, Tong and Jiaye, Ge and Li, Mo and Liu, Hongwei and Liu, Junnan and Liu, Yuan and Lyu, Chengqi and Lyu, Han and Ma, Ningsheng and Ma, Zerun and Sun, Yu and Wu, Zhiyong and Xiao, Linchen and Xiong, Zhuozhi and Xu, Jun and Ye, Haochen and Yu, Zhaohui and Yuan, Yike and Zhang, Songyang and Zhao, Yufeng and Zhou, Fengzhe and Zhou, Peiheng and Zhu, Dongsheng and Zhu, Lin and Zhuo, Jingming},
  year = {2023},
  eprint = {2605.19276},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
  url = {https://arxiv.org/abs/2605.19276},
  journal = {arXiv Preprint}
}