Research map / Probabilistic and scientific modeling / Molecules and proteins
ContinuAR: Continuous Autoregression For Infinite-Fidelity Fusion
WEI W. XING, Yuxin Wang, Zheng Xing
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 |
|---|---|---|---|
| ContinuAR (continuous fidelity indicator) vs best competitor on subset training | ContinuAR · Heat/Burger's/Poisson, subset | RMSE: 3.5x, 2.5x, 8.7x accuracy improvement | Improved |
| non-subset training data (no subset structure across fidelities) | ContinuAR · Heat/Burger's/Poisson, non-subset | RMSE: 1.9x, 1.6x, 1.5x accuracy improvements | Improved |
| larger eta (less reduction of high-fidelity samples) (ablation) | all methods · PDE benchmarks | RMSE | Improved |
| complex real-world data (TopOP, PNA) | ContinuAR · TopOP, PNA | RMSE: 1.1x and 1.4x improvements | Improved |
| use ContinuAR instead of IFC for training cost | ContinuAR · Heat, N0=32, eta=0.5 | training time: 603x | Improved |
| use ContinuAR instead of IFC | ContinuAR · Heat, N0=32, eta=0.5 | RMSE: 2.8x | Improved |
| use IFC instead of AR/ResGP on PNA | IFC · PNA subset vs non-subset | RMSE: 192s training, 140x more than ResGP and AR | Mixed |
| use fidelity indicator t_i (continuous fidelity) as in IFC and ContinuAR | fidelity fusion · PDE benchmarks | RMSE | Improved |
| model complexity of MF-BNN | MF-BNN · PDE benchmarks | RMSE | Hurt |
| random-sample cost evaluation with relaxed low-fidelity sample requirement | ContinuAR · Heat/Burger's/Poisson/TopOP/PNA | RMSE vs data cost: 4.3x, 9.3x, 2.7x, 1.2x, and 1.3x | Improved |
Closest papers by meaning
- Compute-optimal data scaling for neural surrogates via multi-fidelity training (NeurIPS 2026 · similarity 0.85)
- Consistency Models Made Easy (ICLR 2025 · similarity 0.85)
- Multi-Fidelity Residual Neural Processes for Scalable Surrogate Modeling (ICML 2024 · similarity 0.85)
- Elucidating the Preconditioning in Consistency Distillation (ICLR 2025 · similarity 0.84)
- Model-Agnostic Knowledge Guided Correction for Improved Neural Surrogate Rollout (ICLR 2025 · similarity 0.83)
- Self-Refining Diffusion Samplers: Enabling Parallelization via Parareal Iterations (NeurIPS 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.
