Table 9 Comparison of interpretability characteristics between TabPFN and M5Prime surrogate models.

From: Machine learning based optimization of fly ash content for improving geopolymer concrete compressive strength

Criterion

TabPFN

M5Prime

Type of interpretability

Post-hoc (e.g., with SHAP)

Intrinsic (conditional rules and linear equations)

Human readability

Lower

Higher

Need for auxiliary tools

Yes

No

Decision path traceability

Indirect

Direct

Stability of feature importance

Dependent on interpretation method

High and intrinsic

Back to article page