Table 3.

Results for BERT and Llama 2 and 3 in the test of $\mathit{A}\cap \mathit{B}\cap \mathit{C}$

	BERT					Llama 2		Llama 3
Cohort	Recall (accuracy)	Precision	Precision (outside)	F1	F1 (outside)	Exact match	80% match	Exact match	80% match
Superhero	0.94	0.55	1.00	0.69	0.97	0.07	0.08	0.10	0.12
Dinosaur	0.96	0.54	0.98	0.69	0.97	0.10	0.12	0.10	0.12
Mammal	1.00	0.56	0.93	0.72	0.96	0.14	0.14	0.11	0.15
Bird	0.96	0.53	0.96	0.68	0.96	0.08	0.10	0.09	0.10

“Precision (outside)” means that negative samples contain names outside the union of A, B, and C. “F1(outside)” means the F1 score between Recall and Precision(outside).