Abstract: To accurately predict the lower heating value (LHV) of municipal solid waste (MSW) fed into waste-to-energy incineration, we compiled data from nine published studies and 40 in-house experimental records to construct a dataset of 277 samples. Seven MSW component fractions and moisture content were used as input features. A small-sample Transformer model, Tabular Prior-Data Fitted Network (TabPFN), was introduced for LHV prediction and benchmarked against mainstream models such as support vector regression (SVR), random forest (RF), and CatBoost. In addition, the accumulated local effects (ALE) method was employed to interpret the governing factors and elucidate their underlying influence patterns from both univariate marginal and multivariate interaction perspectives. Results show that TabPFN achieves the highest predictive accuracy under limited data (R²=0.863), without the need for hyperparameter tuning. It requires a total prediction time of 3.27 s, which is more than two orders of magnitude faster than CatBoost at comparable accuracy, thus delivering both high accuracy and computational efficiency for small samples. ALE analysis identifies moisture content and inorganic fraction as the primary factors suppressing LHV, while plastics/rubber and paper are the key combustible components for LHV enhancement, with all components exhibiting significant nonlinear threshold effects: suppression intensifies markedly when moisture content exceeds 55% and inorganic fraction exceeds 5%, while the positive effect of plastics/rubber becomes particularly prominent when its content exceeds 20%. Interaction analysis further reveals that moisture content is a critical prerequisite for combustible components to effectively contribute to heat generation: below 55%, plastics/rubber and paper exhibit significant synergistic heating effects; once exceeded, the marginal contributions of combustible fractions are substantially suppressed. From an operational perspective, priority should be given to controlling moisture content below 55% in MSW pretreatment, followed by moderately increasing the proportions of plastics/rubber and paper to further improve LHV of waste and operational stability of the incineration system.