AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Rethinking BCE Loss for Multi-Label Image Recognition with Fine-Tuning

Ao Zhou, Zhiwei Jiang, Zifeng Cheng, Cong Wang, Yafeng Yin, Shufan Yang, Qing Gu

CVPR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Parameter-efficient visual tuning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add class-wise covariance regularization (CCR) to BCE fine-tuningCoOp · Six datasets avgECE (x10^-2): 13.25 -> 7.35Improved
add CCR to BCE fine-tuningCoOp · New classes (six datasets avg)ECE (x10^-2): 21.58 -> 11.03Improved
add CCR to BCE fine-tuningT2I-PAL · Six datasets avgECE (x10^-2): 4.09 -> 3.62Improved
add CCR to DualCoOpDualCoOp · Six datasets avgECE (x10^-2): 4.53 -> 4.57No clear effect
add CCR to TaI-DPTTaI-DPT · COCO-LT target (domain generalization)ECE (x10^-2): 5.23 -> 5.58Hurt
add CCR to CoOpCoOp · MS-COCO source accuracyaccuracy: 69.44 -> 71.47Improved
add CCR to TaI-DPTTaI-DPT · COCO-2014 target accuracyaccuracy: 69.57 -> 74.94Improved
apply DAC or DOR calibration instead of CCR to CoOpCoOp · Tail classesECE (x10^-2): 8.37 -> 9.28 (DAC)Mixed
vary regularization coefficient lambda (ablation)CCR · base classes, CoOpaccuracy and ECENo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy