Research map / Label-efficient and robust learning / Recognition and long-tail
Agreement aware and dissimilarity oriented GLOM
Ru Zeng, Yan Song, Yang Zhang, Yanling Hu, Hui Yu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace full-resolution top-level heads with the dissimilarity-focused head Hd (ablation) | GLOM variant on CIFAR-10 · CIFAR-10 | params (millions) and error %: 72 -> 2 params; error 9.1 -> 11.6 | Mixed |
| add contrastive agreement enhancer (CAE) without Hd (ablation) | GLOM baseline · CIFAR-10 | error %: 9.1 -> 9.0 | No clear effect |
| add CAE on top of Hd (ablation) | Hd model · CIFAR-10 | error %: 11.6 -> 10.9 | Improved |
| use Hd instead of average or max pooling for top-level downsampling (ablation) | GLOM variant · CIFAR-10 | error %: 11.1 -> 10.9 | Improved |
| add CAE to improve agreement versus the Agglomerator implementation | GLOM · CIFAR-10 to F-MNIST, TVn | total variance of normalized embeddings | Improved |
| use CAE + Hd model instead of Agglomerator on CIFAR-100 | GLOM · CIFAR-100 | error %: 40.97 -> 41.31 | Hurt |
| use CAE + Hd model instead of Agglomerator on Fashion-MNIST | GLOM · Fashion-MNIST | error %: 7.43 -> 5.10 | Improved |
Closest papers by meaning
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.83)
- Diversity Is All You Need for Contrastive Learning: Spectral Bounds on Gradient Magnitudes (NeurIPS 2025 · similarity 0.83)
- The Loss Is Not Enough: Sampling Conditions and Inductive Bias in Contrastive Representation Learning (ICML 2026 · similarity 0.82)
- Contrasting Quadratic Assignments for Set-Based Representation Learning (ECCV 2022 · similarity 0.82)
- Variational Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.82)
- ACCon: Angle-Compensated Contrastive Regularizer for Deep Regression (AAAI 2025 · similarity 0.82)
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.
