Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Interpretable Compositional Representations for Robust Few-Shot Generalization
Samarth Mishra, Pengkai Zhu, Venkatesh Saligrama
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 |
|---|---|---|---|
| add pseudo-labeled target images (joint pi) to RPC training | RPC (source pi) · SVHN->MNIST DA | accuracy: 82.4 -> 95.2 | Improved |
| add pseudo-labeled target images (joint pi) to RPC training | RPC (source pi) · MNIST->USPS DA | accuracy: 94.8 -> 98.8 | Improved |
| add pseudo-labeled target images (joint pi) to RPC training | RPC (source pi) · USPS->MNIST DA | accuracy: 96.1 -> 96.8 | Improved |
| use RPC encoding with source-only training instead of a standard DCN | Source Only · SVHN->MNIST DA | accuracy: 60.1 -> 82.4 | Improved |
| add compactness loss L_com (ablation) | RPC · CUB 5-way 1-shot | accuracy: 73.01 -> 75.01 | Improved |
| add diversity loss L_div (ablation) | RPC · CUB 5-way 1-shot | accuracy: 74.72 -> 75.01 | Improved |
| add autoencoder loss L_ae (ablation) | RPC · CUB 5-way 1-shot | accuracy: 71.84 -> 75.01 | Improved |
| add autoencoder loss L_ae (ablation) | RPC · CUB 5-way 5-shot | accuracy: 85.63 -> 89.61 | Improved |
| add compactness loss L_com (ablation) | RPC · GZSL CUB | harmonic mean: 39.6 -> 48.4 | Improved |
| add diversity loss L_div (ablation) | RPC · GZSL CUB | harmonic mean: 43.7 -> 48.4 | Improved |
| add autoencoder loss L_ae (ablation) | RPC · GZSL CUB | harmonic mean: 48.1 -> 48.4 | No clear effect |
| add autoencoder loss L_ae (ablation) | RPC · SVHN->MNIST DA | accuracy: 94.4 -> 95.2 | Improved |
Closest papers by meaning
- Provable Compositional Generalization for Object-Centric Learning (ICLR 2024 · similarity 0.88)
- Unlocking Compositional Generalization in Continual Few-Shot Learning (NeurIPS 2026 · similarity 0.88)
- Beyond Image Classification: A Video Benchmark and Dual-Branch Hybrid Discrimination Framework for Compositional Zero-Shot Learning (CVPR 2025 · similarity 0.87)
- A Dynamic Learning Method towards Realistic Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.87)
- Learning to Compose: Improving Object Centric Learning by Injecting Compositionality (ICLR 2024 · similarity 0.87)
- Does Data Scaling Lead to Visual Compositional Generalization? (ICML 2025 · similarity 0.87)
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.
