Question Answering on RAG Benchmarks benchmark leaderboard · SOTA2 Research

Question Answering on RAG Benchmarks

63.82HotpotQA ScoreFull-FT

Evaluation Results

Method	Links
Full-FT2026.04		63.82	71.38	75.83	24.2	53.56	57.76
GSA2026.04		59.88	68.62	71.99	20.27	46.2	53.39
H-GSA2026.04		56.43	66.76	69.56	18.37	42.48	50.72
GSA2026.04		54.29	66.16	67.93	16.3	41.78	49.29
UniGist2026.04		49.25	64	64.17	12.37	36.56	45.27
UniGist2026.04		43.5	61.83	60.72	10.88	34.26	42.24
KVLink2026.04		41.15	61.65	62.68	10.47	31.28	41.45
GSA2026.04		41.09	52.06	38.65	12.99	23.62	33.68
KVLink2026.04		36.92	58.85	62.21	8.11	27.24	38.67
H-GSA2026.04		34.34	50.98	33.61	9.06	22.98	30.19
GSA2026.04		30.56	46.28	32.67	8.11	21.1	27.74
Llama3.2-1b2026.04		29.98	40.82	25.86	9.1	34.2	27.99
Full-PT2026.04		29.64	39.73	25.45	9.18	31.72	27.14
UniGist2026.04		22.77	44.44	22.84	3.72	18.86	22.53
UniGist2026.04		21.51	41.62	30.1	3.23	14.56	22.2
KVLink2026.04		17.54	44.36	25.03	4.01	16.96	21.58
KVLink2026.04		10.67	33.1	21.22	0.74	4.92	14.13