Beyond Accuracy: What Matters in Designing Well-Behaved Image Classification Models?
超越准确率:设计良好行为图像分类模型所关注的问题
AI总结 本文研究了图像分类模型在鲁棒性、校准和公平性等质量维度上的表现,提出QUBA指标以评估和推荐模型。
Comments Published in TMLR (01/2026) | OpenReview: https://openreview.net/forum?id=E7HDtLCoT6 | Project page: https://visinf.github.io/beyond-accuracy/