AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Enhanced Multi-Instance Partial Label Learning via Average Gradient Outer Product

Nan Cao, Xu Zhao, Teng Zhang

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Partial and multi-label learning

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 AGOP feature rescaling to the attention module (ablation)AGOPMIPL · MNIST-MIPL (r=3)accuracy: .891 -> 1.00Improved
add AGOP feature rescaling to the attention module (ablation)AGOPMIPL · FMNIST-MIPL (r=3)accuracy: .621 -> .752Improved
add AGOP feature rescaling to the attention module (ablation)AGOPMIPL · CRC-MIPL-KMeansSeg (r=3)accuracy: .618 -> .722Improved
add progressive label disambiguation (ablation)AGOPMIPL · FMNIST-MIPL (r=3)accuracy: .698 -> .752Improved
add progressive label disambiguation (ablation)AGOPMIPL · MNIST-MIPL (r=3)accuracy: .985 -> 1.00Improved
aggregate raw instance features instead of AGOP-transformed features for the bag representation (ablation)AGOPMIPL · Appendix BaccuracyImproved
vary number of prototypes np in {q/2, q, 2q} (ablation)AGOPMIPL · SIVAL-MIPLaccuracy: less than 1.1 pointsNo clear effect
vary loss weights mu and gamma (ablation)AGOPMIPL · FMNIST-MIPL (r=1), CRC-MIPL-SIFTaccuracyNo clear effect
use AGOPMIPL instead of FastMIPL under heavy candidate noiseAGOPMIPL · SIVAL-MIPL (r=5)accuracy: .449 -> .544Improved
use AGOPMIPL instead of FastMIPL on Birdsong-MIPL at higher noiseAGOPMIPL · Birdsong-MIPL (r=3)accuracy: .767 -> .742Hurt
increase candidate false-positive count r from 1 to 3baselines such as ProMIPL · MNIST-MIPLaccuracy: .999 -> .783Hurt

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