Rethinking Classifier Re-Training in Long-Tailed Recognition: Label Over-Smooth Can Balance

Siyu Sun, Han Lu, Jiangtong Li, Yichen Xie, Tianjiao Li +3 more
2/12/2026

No official code found. Best implementation: Thinklab-SJTU/LOS (9 stars, MIT, last commit 2025-02-14).

Semantic Scholar

Code Implementations(1)

Thinklab-SJTU/LOSCommunity74%
MIT

Official Implementation of Our ICLR 2025 Paper "Rethinking Classifier Re-Training in Long-Tailed Recognition: Label Over-Smooth Can Balance"

91PythonFeb 11, 20251 years agoMIT

Cite this paper

@article{sun2026rethinking,
  title  = {Rethinking Classifier Re-Training in Long-Tailed Recognition: Label Over-Smooth Can Balance},
  author = {Siyu Sun and Han Lu and Jiangtong Li and Yichen Xie and Tianjiao Li and Xiaokang Yang and Liqing Zhang and Junchi Yan},
  year   = {2026},
  url    = {https://api.semanticscholar.org/CorpusID:a20cb7364921e79c517f80765d1723df37701120},
  journal = {ICLR 2025 2025}
}

Discussion