Hi, My name is Haobin Li. I’m a Ph.D. student from College of Computer Science, Sichuan Univerisity, fortunately advised by Prof. Xi Peng.

My research interests mainly focus on achieving robustness against open-world challenges in Multi-modal Learning and Agentic LLM.

🔥 News

  • 2026.08: One paper was accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI 2026).
  • 2026.05: One paper was accepted by International Conference on Machine Learning (ICML 2026).
  • 2026.01: One paper was accepted by International Conference on Learning Representations (ICLR 2026).

📝 Publications

* Equal contribution Corresponding author

arXiv
sym

[arXiv’26] Active-SWE: Benchmarking Coding Agents for Proactive Bug Fixing without Issue Reports
Haobin Li, Ping Deng, Weizhong Qian, Liang Jiang, Zhenyu Huang, Mouxing Yang, Xi Peng

Code | Hugging Face | Homepage / Leaderboard | 中文简介 |

  • The first benchmark that evaluates coding agents in proactive bug fixing.
  • 1,663 high-quality tasks across six bug categories and eight languages.
ICML
sym

[ICML’26] Reliable Thinking with Images
Haobin Li, Yutong Yang, Yijie Lin, Xiang Dai, Mouxing Yang, Xi Peng

Code | PaperWeekly 报道 | AI TIME 报道 |

  • First reveal and study noisy mining and reasoning in thinking with images.
  • Pioneering exploration of reliable multimodal reasoning.
IEEE TPAMI
sym

[TPAMI’26] Toward Robust and Harmonious Adaptation for Cross-modal Retrieval
Haobin Li, Mouxing Yang, Xi Peng

Code |

  • Extend query shift and tackle more challenging diverse shift scenarios.
  • Establish more benchmarks across image-text, video–audio and composed image retrieval.
ICLR Oral
sym

[ICLR’26 Oral] Learning with Dual-level Noisy Correspondence for Multi-modal Entity Alignment
Haobin Li, Yijie Lin, Peng Hu, Mouxing Yang, Xi Peng

Code | 机器之心报道 |

  • First reveal and study noisy and underlying correspondence in multi-modal knowledge graph.
  • Pioneering exploration of test-time robustness.
IEEE TPAMI
sym

[TPAMI’26] Community-aware Multi-view Representation Learning with Incomplete Information
Haobin Li, Yijie Lin, Peng Hu, Mouxing Yang, Xi Peng

Code |

  • Leveraging sociological concepts to address missing modalities and correspondences.
ICLR Spotlight
sym

[ICLR’25 Spotlight] Test-time Adaptation for Cross-modal Retrieval with Query Shift
Haobin Li, Peng Hu, Qianjun Zhang, Xi Peng, Xiting Liu, Mouxing Yang

Code | 机器之心报道 |

  • First study the query shift problem in cross-modal retrieval.
  • Reveal the underlying impacts of query shift on cross-modal retrieval
IJCAI
sym

[IJCAI’23] Incomplete Multi-view Clustering via Prototype-based Imputation
Haobin Li, Yunfan Li, Mouxing Yang, Peng Hu, Dezhong Peng, Xi Peng

Code |

  • Leveraging sample-prototype relationship to impute missing modality.
Vicinage Earth 2024
sym

[Vicinagearth’24] A survey on deep clustering: from the prior perspective
Yiding Lu*, Haobin Li*, Yunfan Li, Yijie Lin, Xi Peng

中文简介 |

  • A survey of deep clustering from a prior perspective.

📖 Service

  • Journal Reviewer: IEEE TPAMI, IEEE TIP, and more.
  • Conference Reviewer: ICLR, NeurIPS, ICML, CVPR, AAAI, and more.

💻 Internships

  • 2026.01 - Now, Ant Group, Ling Team: Post Training.