Table 17. Benchmark of the different LLMs with the whole corpus and the two splits evaluated for multi-label sentiment classification.

For each model and dataset, weighted precision (W-P), weighted recall (W-R), weighted (W-F1) and macro (M-F1) are reported.

Dataset	Model	W-P	W-R	W-F1	M-F1
Tweets	BETO	0.7476	0.7362	0.7314	0.6397
	ALBETO	0.7170	0.6997	0.6928	0.5664
	DistilBETO	0.7060	0.7058	0.6971	0.5729
	MarIA	0.7669	0.7445	0.7395	0.6398
	BERTIN	0.7421	0.7171	0.7102	0.6079
Headlines	BETO	0.7400	0.7253	0.7284	0.6515
	ALBETO	0.7137	0.7006	0.7021	0.6133
	DistilBETO	0.6973	0.6710	0.6808	0.5892
	MarIA	0.7439	0.7266	0.7301	0.6542
	BERTIN	0.7301	0.7188	0.7165	0.6171
Total	BETO	0.7734	0.7640	0.7680	0.7113
	ALBETO	0.7399	0.7332	0.7324	0.6543
	DistilBETO	0.7468	0.7279	0.7338	0.6561
	MarIA	0.7827	0.7597	0.7678	0.7017
	BERTIN	0.7600	0.7553	0.7516	0.6679