Research map / Label-efficient and robust learning / Self-supervised and contrastive
Robust Self-Supervised Multi-Instance Learning with Structure Awareness
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use SMILES instead of the strongest reported baseline | MIL classification · Five MIL benchmarks (average) | accuracy (%): 87.4±3.7 -> 92.9±2.1 | Helped |
| Use SMILES instead of the strongest reported baseline | 20 Newsgroups MIL classification · 20 Newsgroups (average) | accuracy (%): 81.7±3.2 -> 87.2±2.8 | Helped |
| Add bag augmentation (ablation) | SMILES · MUSK2, FOX, TIGER, ELEPHANT | accuracy (%): 88.3±1.7 -> 89.9±3.2; 80.5±1.3 -> 85.3±3.2; 90.1±2.9 -> 91.1±2.4; 92.2±2.3 -> 96.8±2.0 | Helped |
| Add a learned bag structure graph (ablation) | SMILES · MUSK2 | accuracy (%): 80.2±1.4 -> 95.5±2.1 | Helped |
| Add a learned bag structure graph (ablation) | SMILES · FOX | accuracy (%): 65.8±2.1 -> 85.1±1.3 | Helped |
| Add a learned bag structure graph (ablation) | SMILES · TIGER | accuracy (%): 62.1±4.2 -> 91.2±1.2 | Helped |
| Add a learned bag structure graph (ablation) | SMILES · ELEPHANT | accuracy (%): 69.2±3.6 -> 97.5±1.5 | Helped |
| Increase augmentation ratio beyond moderate levels (ablation) | SMILES · FOX | accuracy (%): Mask: 85.3±3.2 -> 81.3±1.6 (20% -> 50%); Drop: 83.2±2.8 -> 73.7±1.4 (20% -> 50%) | Hurt |
| Train with contrastive bag augmentation under injected noise | SMILES · COMPONENT, FUNCTION, PROCESS | accuracy (%): COMPONENT η=0.0: 91.1 ± 1.5; η=0.4: 89.1 ± 4.1; FUNCTION η=0.0: 94.7 ± 2.7; η=0.4: 92.7 ± 1.2; PROCESS η=0.0: 96.2 ± 3.6 | Mixed |
Closest papers by meaning
- Rethinking Multi-Instance Learning Through Graph-Driven Fusion: A Dual-Path Approach to Adaptive Representation (AAAI 2026 · similarity 0.91)
- Bag Graph: Multiple Instance Learning Using Bayesian Graph Neural Networks (AAAI 2022 · similarity 0.89)
- Multi-Instance Causal Representation Learning for Instance Label Prediction and Out-of-Distribution Generalization (NeurIPS 2022 · similarity 0.88)
- Sm: enhanced localization in Multiple Instance Learning for medical imaging classification (NeurIPS 2024 · similarity 0.84)
- Equivalence is All: A Unified View for Self-supervised Graph Learning (ICML 2025 · similarity 0.84)
- Are Multiple Instance Learning Algorithms Learnable for Instances? (NeurIPS 2024 · similarity 0.84)
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.