Table 5.

Severity prediction performance.

Model	Model category	MAE ↓	RMSE ↓	Spearman ρ ↑
BERT regression	General-domain transformer	3.52	4.67	0.61
RoBERTa regression	General-domain transformer	3.31	4.42	0.64
MentalRoBERTa (Ji et al., 2022)	Mental-health-adapted transformer	3.05	4.06	0.69
Prompt-only LLM GPT-5.0	Foundation-model baseline	3.27	4.35	0.65
Evidence linear model	Evidence-based baseline	3.12	4.10	0.68
R-GCRED dual-h.ead	Retrieval-grounded calibrated framework	2.78	3.82	0.74