[12] Spectral Imbalance Causes Forgetting in Low-Rank Continual Adaptation [pdf] [code]
Hao Gu, Mao-Lin Luo Zi-Hao Zhou, Han-Chen Zhang, Min-Ling Zhang, Tong Wei†.
In: Proceedings of the 43rd International Conference on Machine Learning (ICML'26), Seoul, South Korea, 2026.

[11] DC-Merge: Improving Model Merging with Directional Consistency [pdf] [code]
Han-Chen Zhang*, Zi-Hao Zhou*, Mao-Lin Luo, Shimin Di, Min-Ling Zhang, Tong Wei†.
In: Proceedings of the 37th IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR'26), Denver CO, 2026.

[10] KeepLoRA: Continual Learning with Residual Gradient Adaptation [pdf] [code]
Mao-Lin Luo, Zi-Hao Zhou, Yi-Lin Zhang, Yuanyu Wan, Min-Ling Zhang, Tong Wei†.
In: Proceedings of the 14th International Conference on Learning Representations (ICLR'26), Rio de Janeiro, Brazil, 2026.

[9] X-Mahalanobis: Transformer Feature Mixing for Reliable OOD Detection [pdf] [code]
Tong Wei, Bo-Lin Wang, Jiang-Xin Shi, Yu-Feng Li, Min-Ling Zhang
In: Advances in Neural Information Processing Systems 38 (NeurIPS'25), San Diego, USA, 2025.

[8] LADA: Scalable Label-Specific CLIP Adapter for Continual Learning [pdf] [code]

Mao-Lin Luo, Zi-Hao Zhou, Tong Wei†, Min-Ling Zhang.
In: Proceedings of the 42nd International Conference on Machine Learning (ICML'25), Vancouver, Canada, 2025.

[7] Weakly-Supervised Contrastive Learning for Imprecise Class Labels [pdf] [code]
Zi-Hao Zhou, Jun-Jie Wang, Tong Wei†, Min-Ling Zhang.
In: Proceedings of the 42nd International Conference on Machine Learning (ICML'25, Spotlight), Vancouver, Canada, 2025.

[6] Semi-Supervised CLIP Training by Enforcing Semantic and Trapezoidal Consistency [pdf] [code]
Kai Gan, Bo Ye, Min-Ling Zhang, Tong Wei†.
In: Proceedings of the 13th International Conference on Learning Representations (ICLR'25), Singapore, 2025.

[5] Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition [pdf] [code]
Zi-Hao Zhou*, Siyuan Fang*, Zi-Jing Zhou, Tong Wei†, Yuanyu Wan, Min-Ling Zhang.
In: Advances in Neural Information Processing Systems 37 (NeurIPS'24), Vancouver, Canada, 2024, 51411-51435.

[4] Vision-Language Models are Strong Noisy Label Detectors [pdf] [code]
Tong Wei, Hao-Tian Li, Chun-Shu Li, Jiang-Xin Shi, Yu-Feng Li, Min-Ling Zhang.
In: Advances in Neural Information Processing Systems 37 (NeurIPS'24), Vancouver, Canada, 2024, 58154-58173.

[3] Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning [pdf] [code]
Kai Gan, Tong Wei†
In: Proceedings of the 41st International Conference on Machine Learning (ICML'24), Vienna, Austria, 2024, 14453-14470.

[2] Learning Label Shift Correction for Test-Agnostic Long-Tailed Recognition [pdf] [code]
Tong Wei, Zhen Mao, Zi-Hao Zhou, Yuanyu Wan, Min-Ling Zhang
In: Proceedings of the 41st International Conference on Machine Learning (ICML'24), Vienna, Austria, 2024, 52611-52631.

[1] Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts [pdf] [code]
Jiang-Xin Shi*, Tong Wei*, Zhi Zhou, Xin-Yan Han, Jie-Jing Shao, Yu-Feng Li
In: Proceedings of the 41st International Conference on Machine Learning (ICML'24), Vienna, Austria, 2024, 45014-45039