OSDTW: Optimal Shared Depth and Task Weighting for Long-Tailed Recognition 事件

OSDTW: Optimal Shared Depth and Task Weighting for Long-Tailed Recognition arXiv:2605.24969v1 Announce Type: cross Abstract: Long-tailed recognition suffers from a persistent head--tail trade-off: improving tail performance often degrades head accuracy and can increase training instability. Despite strong empirical results from re-weighting, decoupled training, and multi-expert methods, key design choices about representation sharing between head and tail classes and supervision weighting acros