Research map / Label-efficient and robust learning / Recognition and long-tail
Credible Information Subset Decomposition: An End-to-End Multi-fidelity Learning Model by Modeling Label Information
Sihan Wang, Wenjie Du, Yang Wang
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 evidence regression (replacing MSE) for absolute-value labels (ablation) | Credible Subset Decomposition · TF-Bandgap (2-fidelity) | MAE: 0.66 -> 0.57 | Improved |
| add interval-based uncertainty modulation with tau (ablation) | Credible Subset Decomposition · TF-Bandgap (2-fidelity) | MAE: 0.60 -> 0.57 | Improved |
| add ranking feature and rank loss (ablation) | Credible Subset Decomposition · FF-Bandgap (5-fidelity) | Kendall tau-b: 0.64 -> 0.70 | Improved |
| add ranking feature and rank loss (ablation) | Credible Subset Decomposition · TF-Bandgap (2-fidelity) | MAE: 0.58 -> 0.57 | Improved |
| add interval-based uncertainty modulation with tau (ablation) | Credible Subset Decomposition · FF-Bandgap (5-fidelity) | MAE: 0.55 -> 0.50 | Improved |
| use credible subset decomposition instead of best multi-fidelity baseline MFSE | multi-fidelity learning · TF-Bandgap (2-fidelity) | MAE: 0.64 -> 0.57 | Improved |
| use credible subset decomposition instead of MFGNN with five fidelities | multi-fidelity learning · FF-Bandgap (5-fidelity) | MAE: 0.56 -> 0.50 | Improved |
| add a fourth fidelity level (GLLB-SC data) to training mix (ablation) | Credible Subset Decomposition · FF-Bandgap HSE/SCAN/PBE test sets | MAE | Hurt |
| add more high-fidelity data to the training mix (ablation) | Credible Subset Decomposition · FF-Bandgap | MAE | Improved |
| reduce alpha_abs from 1.0 to 0.1 (rank weight fixed 5e-3) (ablation) | Credible Subset Decomposition · TF-Bandgap (2-fidelity) | MAE: 0.57 -> 0.69 | Hurt |
| increase alpha_rank from 1e-3 to 1e-1 (alpha_abs fixed 1.0) (ablation) | Credible Subset Decomposition · TF-Bandgap (2-fidelity) | MAE: 0.58 -> 0.68 | Hurt |
Closest papers by meaning
- Chem-PerturBridge: a harmonized compendium of small molecule perturbation transcriptomic effects (NeurIPS 2026 · similarity 0.85)
- Instructor-inspired Machine Learning for Robust Molecular Property Prediction (NeurIPS 2024 · similarity 0.84)
- ChemPile: A 250 GB Diverse and Curated Dataset for Chemical Foundation Models (NeurIPS 2025 · similarity 0.83)
- MS-BART: Unified Modeling of Mass Spectra and Molecules for Structure Elucidation (NeurIPS 2025 · similarity 0.83)
- Physical Consistency Bridges Heterogeneous Data in Molecular Multi-Task Learning (NeurIPS 2024 · similarity 0.83)
- Learning Molecular Representation in a Cell (ICLR 2025 · 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.
