Research map / Large language models / AI and society
SMRS: advocating a unified reporting standard for surrogate models in the artificial intelligence era.
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Position paper proposing the Surrogate Model Reporting Standard (SMRS), a modular checklist-style schema for documenting data generation, model class, learning method, and evaluation in surrogate modelling, illustrated with case-study audits of published surrogates.
Problem
Surrogate modelling lacks standardised reporting across data sampling, model selection, evaluation, and downstream analysis, which limits reproducibility and cross-domain transfer, especially as AI-driven surrogates proliferate.
Approach
Proposes a lightweight, modular, model-agnostic reporting schema (SMRS) with per-stage checklists covering data mode and sampling, fidelity, deterministic vs probabilistic outputs, learning paradigm, and diagnostic and task-based evaluation, and audits published surrogates against it.
Key results
- Audit of 17 recent surrogate modelling studies found substantial variation in what is reported; few consistently report data collection, sampling decisions, uncertainty quantification, or limitations.
- Four case-study papers (storm surge, geophysical flow, neural fiber, deep adaptive sampling) were mapped onto the SMRS schema, with gaps marked by check/question/cross symbols.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the AxonML surrogate instead of single-core NEURON simulations | MRG axon simulation · MRG axon simulations | speedup: 2,000 to 130,000× speedup | Helped |
| Deploy the surrogate model on GPUs | AxonML surrogate · GPU deployment | speedup: 5 orders-of-magnitude speedup | Helped |
| Use deep adaptive sampling | Surrogate models for parametric differential equations · Physics-informed operator learning, parametric optimal control, and parametric lid-driven 2D cavity flow | speedup: speed ups of order exceeding $10^4$ | Helped |
Surprising findings
- Several published surrogates do not compare their computational gains against the expensive simulator they replace, or omit basic training details such as optimiser and epochs.
Limitations
- Schema is proposed but not empirically validated with authors or reviewers beyond a small case-study audit.
- Case-study selection from 17 papers is ad hoc and assessments rely on the authors' interpretation.
- Adoption depends on incentives and integration with legacy and proprietary simulation tools.
Experimental setup
Builds on: model cards FAIR principles simulation-based calibration physics-informed neural networks
Closest papers by meaning
- Model-Free Assessment of Simulator Fidelity via Quantile Curves (NeurIPS 2026 · similarity 0.84)
- Interventionally Consistent Surrogates for Complex Simulation Models (NeurIPS 2024 · similarity 0.84)
- AstaBench: Rigorous Benchmarking of AI Agents with a Scientific Research Suite (ICLR 2026 · similarity 0.83)
- MLR-Bench: Evaluating AI Agents on Open-Ended Machine Learning Research (NeurIPS 2025 · similarity 0.83)
- Model-Agnostic Knowledge Guided Correction for Improved Neural Surrogate Rollout (ICLR 2025 · similarity 0.83)
- Rigor in AI: Doing Rigorous AI Work Requires a Broader, Responsible AI-Informed Conception of Rigor (NeurIPS 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.