Research map / Architectures and efficiency / Neural architecture search
NAS-Bench-Suite-Zero: Accelerating Research on Zero Cost Proxies
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 |
|---|---|---|---|
| add 13 ZC proxies as features to the surrogate | XGBoost predictor · NAS-Bench-201 CIFAR-100 | Spearman rank correlation: 0.640 -> 0.908 | Helped |
| remove architecture encoding, keep only ZC proxy features (ablation) | XGBoost with ZC proxies · NAS-Bench-201 CIFAR-100 | Spearman rank correlation: 0.908 -> 0.907 | No clear effect |
| debias with performance-tuned C | l2-norm · NB201-CF10 | Pearson correlation with accuracy: 0.42 -> 0.44 | Helped |
| debias with performance-tuned C | synflow · NB201-CF100 | Pearson correlation with accuracy: 0.68 -> 0.71 | Helped |
| debias with performance-tuned C | synflow · NB201-IM | Pearson correlation with accuracy: 0.76 -> 0.76 | No clear effect |
| debias with performance-tuned C | nwot · NB301-CF10 | Pearson correlation with accuracy: 0.49 -> 0.49 | No clear effect |
| fully minimize bias to 0.00 (ablation) | l2-norm · NB201-CF10 | Pearson correlation with accuracy: 0.42 -> 0.10 | Hurt |
| fully minimize bias to 0.00 (ablation) | nwot · NB301-CF10 | Pearson correlation with accuracy: 0.49 -> 0.03 | Hurt |
| fully minimize bias (ablation) | synflow · NB201-IM | Pearson correlation with accuracy: 0.76 -> 0.62 | Hurt |
| equalize bias with val-acc bias (ablation) | flops · NB301-CF10 | Pearson correlation with accuracy: 0.43 -> -0.05 | Hurt |
| add ZC proxy features to surrogate in BANANAS / NPENAS | BANANAS · NAS-Bench-201 | validation accuracy over search | Helped |
Closest papers by meaning
- Zero-Cost Proxies for Lightweight NAS (ICLR 2021 · similarity 0.93)
- ParZC: Parametric Zero-Cost Proxies for Efficient NAS (AAAI 2025 · similarity 0.92)
- AZ-NAS: Assembling Zero-Cost Proxies for Network Architecture Search (CVPR 2024 · similarity 0.90)
- Surprisingly Strong Performance Prediction with Neural Graph Features (ICML 2024 · similarity 0.89)
- L-SWAG: Layer-Sample Wise Activation with Gradients Information for Zero-Shot NAS on Vision Transformers (CVPR 2025 · similarity 0.88)
- NEAR: A Training-Free Pre-Estimator of Machine Learning Model Performance (ICLR 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.