Zero-shot Reasoning on Language Task Suite (COPA, HS, LB, OBQA, PIQA, Race, SciQ, ARC, SIQA, WG) benchmark leaderboard · SOTA2 Research

Zero-shot Reasoning on Language Task Suite (COPA, HS, LB, OBQA, PIQA, Race, SciQ, ARC, SIQA, WG)

68COPALoopFormer

Evaluation Results

Method	Links
LoopFormer2026.02		68	31	32.35	25.4	63.06	28.23	66.3	31.78	37.77	53.43	43.73
Base2026.02		67	32.72	37.78	26.2	64.69	29.86	69.1	32.29	38.38	51.3	44.93
Naive-Loop-EE2026.02		66	29.68	31.52	27	62.24	28.33	65.5	31.64	36.64	50.59	42.91
Base-Loop-EE-Cons2026.02		66	30.54	32.19	26.8	62.13	28.23	64.9	31.45	37.4	52.88	43.25
LoopFormer2026.02		66	32.3	38.27	26.8	63.33	30.81	68	32.71	37.97	51.94	44.81
TMLT2026.02		65	32.34	39.06	27.8	63.11	29.67	69.8	31.67	36.95	51.54	44.69
TMLT-EE2026.02		65	31.03	35.14	28.4	63.11	28.8	68.8	31.74	37.41	50.83	44.03
Naive-Loop-EE2026.02		65	29.35	31.28	26	61.75	28.32	64.4	31.15	36.8	51.85	42.59
Base2026.02		64	30.45	33.45	25.6	62.51	28.04	67.5	31.07	36.18	48.54	42.73
Base-Loop-EE-Cons2026.02		63	29.72	25.6	27	61.26	28.04	58.4	30.06	36.23	51.7	41.1
Naive-Loop-EE2026.02		63	28.64	27.28	24.6	61.48	26.41	62.3	29.91	35.41	50.67	40.97
LoopFormer2026.02		63	28.69	26.14	26.6	60.1	25.74	58.6	29.29	35.26	50.2	40.36
Base2026.02		61	35.04	41.96	27.6	66	29	70.1	33.43	38.18	50.4	45.27
Base-Loop2026.02		61	30.46	34.68	27	63.22	28.8	63.7	31.71	38.43	49.8	42.88
TMLT-EE2026.02		60	29.88	27.73	26.4	62.02	28.8	61.4	30.59	36.69	51.54	41.5
Base-Loop-EE-Cons2026.02		59	28.28	18.49	25.6	59.52	25.16	53.8	28.94	34.65	52.72	38.62
TMLT-EE2026.02		57	28.18	17.09	25.08	58.68	26.41	50.01	28.26	34.9	50.27	37.59