中文DualMem: 绕过目标性瓶颈实现开放世界目标检测中的校准未知流过滤
ENDualMem: Bypassing the Objectness Bottleneck for Calibrated Unknown-Stream Filtering in Open-World Object Detection
研究揭示开放世界目标检测中未知预测流被背景误报严重污染（占46-71%），真正未知目标不足10%。该问题源于目标性头的信息瓶颈，而非信息缺失。需改进对象性头设计以提升未知检测能力。
arXiv:2605.23634v1 Announce Type: new Abstract: Open-world object detection (OWOD) requires detectors to localize known classes while identifying unknown objects for future incremental learning. We find that the unknown prediction streams of strong OWOD detectors are heavily polluted: on M-OWODB, across PROB, OW-DETR, and HypOW, future-task positive unknowns make up less than 10% of unknown predictions, whereas background false positives account for 46-71%. We show that this is not a missing-information problem, but an information bottleneck at the objectness head. On PROB Task 1, a linear pro