Natural Language Understanding on GLUE RoBERTa LARGE (test dev) benchmark leaderboard · SOTA2 Research

Natural Language Understanding on GLUE RoBERTa LARGE (test dev)

90.57MNLI AccuracyLoRA

Evaluation Results

Method	Links
LoRA2025.12		90.57	90.93	94.78	95.98	90.68	91.39	77.72	64.32	87.04
Full-Adam2025.12		90.51	91.9	94.43	96.44	91.6	92.79	78.81	64.85	87.66
SCALEARN++2023.10		90.31	90.59	94.05	95.93	92.48	88.48	86.28	67.13	88.16
GaLore2025.12		90.27	89.99	94.21	96.1	91.18	90.94	77.11	62.52	86.54
SCALEARNUNIFORM2023.10		90.11	90.05	94.23	95.41	92.11	88.63	84.4	66.98	87.74
SCALEARN2023.10		90.09	90.51	94.18	95.41	92.32	88.09	87.08	65.4	87.91
SCALEARNUNIFORM++2023.10		90.08	90.49	94.12	95.18	92.12	90.05	84.98	64.97	87.75
HYPERFORMER++2023.10		89.79	89.54	93.95	95.22	91.62	88.07	86.28	65.16	87.45
PROPETL2023.10		89.78	89.23	94.32	95.41	91.45	87.65	84.55	65.85	87.28
FOAM2025.12		89.74	89.65	93.88	95.75	92.44	92.33	83.03	64.3	87.64
LORA2023.10		89.66	89.66	94.2	95.47	91.98	87.7	80.51	63.8	86.62
HYPERFORMER2023.10		89.66	90.15	93.95	95.8	91.68	86.6	86.28	61.18	86.91
APOLLO2025.12		89.65	89.27	93.61	95.06	91.66	92.14	80.14	61.13	86.58
ADAPTER2023.10		89.62	89.87	94.13	95.24	91.81	87.82	81.23	64.07	86.72
FINETUNE2023.10		89.57	89.75	93.91	95.3	91.89	86.27	81.52	60.15	86.04
ADAPTERFUSION2023.10		89.57	90.88	94.15	95.87	91.86	88.97	85.7	66.39	87.93
COMPACTER++2023.10		89.15	87.33	92.93	95.41	91.46	87.84	79.71	65.66	86.19
ADAPTER-M2023.10		89.1	89.35	93.64	94.9	88.4	83.09	86.64	56.38	85.19
PROPETL-M2023.10		88.98	89.03	94.14	95.15	91.56	87.83	88.45	60.99	87.01
(IA)32023.10		88.69	87.79	91.72	94.95	91.39	86.37	80.79	64.7	85.8
FINETUNE-M2023.10		87.95	89.82	92.58	94.88	87.04	81.37	84.36	55.32	84.16
ADAPTERSOUP2023.10		65.83	82.37	74.06	93.98	81.67	73.37	67.27	43.7	72.78