Surrogate AI vs classical solvers — validated accuracy and speedup.
Benchmarks compare KYAMOS neural surrogate predictions against full classical numerical solver outputs in curated validation studies.
Benchmark results are curated from KYAMOS validation studies comparing surrogate AI predictions against classical numerical solvers. Selected cases include MSE reduction, heated cavity validation, and inference speedup against full solver workflows.
What it measures: error reduction across surrogate configurations.
Validation context: Measures the reduction in prediction error achieved by the surrogate AI model compared with baseline configurations or classical simulation outputs.
What it measures: thermal-flow prediction accuracy.
Validation context: Validation case for thermal flow behaviour, comparing surrogate predictions against high-fidelity solver results.
What it measures: runtime reduction after training.
Validation context: Runtime comparison between trained neural surrogate inference and full classical solver execution.
Selected supporting publications and technical references are curated on the Research page.