Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Generalizable Neural Fields as Partially Observed Neural Processes
Jeffrey Gu, Kuan-Chieh Wang, Serena Yeung
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 |
|---|---|---|---|
| use PONP neural process instead of Transformer INR hypernetwork | neural field generalization · CelebA image regression | PSNR: 48.36 -> 78.87 | Improved |
| use PONP instead of Transformer INR | neural field generalization · CelebA image completion | PSNR: 18.09 -> 23.24 | Improved |
| add test-time optimization to PONP (ablation) | PONP · CT reconstruction (1 view) | PSNR: 16.39 -> 24.67 | Improved |
| add test-time optimization to PONP (ablation) | PONP · CT reconstruction (8 views) | PSNR: 22.60 -> 37.66 | Improved |
| use PONP with TTO instead of Reptile | CT reconstruction · CT reconstruction (8 views) | PSNR: 27.34 -> 37.66 | Improved |
| use PONP without TTO instead of Reptile with TTO | CT reconstruction · CT reconstruction (8 views) | PSNR: 27.34 -> 22.60 | Hurt |
| use PONP instead of Transformer INR (1 view, no TTO) | view synthesis · ShapeNet chairs | PSNR: 19.66 -> 19.48 | No clear effect |
| use PONP instead of Transformer INR (2 views, no TTO) | view synthesis · ShapeNet cars | PSNR: 25.45 -> 25.98 | Improved |
| use attention-based AttnLNP instead of CNP, no TTO (ablation) | PONP · CT reconstruction (8 views) | PSNR: 22.60 -> 23.88 | Improved |
| use LNP instead of CNP with TTO (ablation) | PONP · CT reconstruction (8 views) | PSNR: 37.66 -> 29.34 | Hurt |
| use AttnCNP instead of CNP with TTO (ablation) | PONP · CT reconstruction (1 view) | PSNR: 24.67 -> 18.32 | Hurt |
| use ConvCNP instead of CNP (ablation) | PONP · CelebA image regression | PSNR: 20.37 -> 78.87 | Improved |
Closest papers by meaning
- Neural Fields as Distributions: Signal Processing Beyond Euclidean Space (CVPR 2024 · similarity 0.84)
- Dimension Agnostic Neural Processes (ICLR 2025 · similarity 0.83)
- How to Train Neural Field Representations: A Comprehensive Study and Benchmark (CVPR 2024 · similarity 0.83)
- Graph Metanetworks for Processing Diverse Neural Architectures (ICLR 2024 · similarity 0.83)
- Generalization Bounds via Meta-Learned Model Representations: PAC-Bayes and Sample Compression Hypernetworks (ICML 2025 · similarity 0.83)
- Sample-Efficient Linear Representation Learning from Non-IID Non-Isotropic Data (ICLR 2024 · similarity 0.83)
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.
