Long-context Understanding on LongBench LLaMA-3.1-8B-Instruct (test) benchmark leaderboard · SOTA2 Research

Long-context Understanding on LongBench LLaMA-3.1-8B-Instruct (test)

32.8NrtvQAStructKV

Evaluation Results

Method	Links
StructKV2026.04		32.8	44.5	55.25	55.8	46.9	30.85	32.1	25.35	25.9	73.65	92.1	43.25	8.85	99.5	63.2	56.8	48.97
StructKV2026.04		32.45	41.88	55.12	55.2	46.85	30.55	29.6	25.15	24.6	73.2	92.45	43	8.6	99.5	62.4	56.15	48.61
SnapKV2026.04		31.51	40.18	41	55.83	44.3	30.92	28.88	24.58	22.57	70.5	91.28	42.64	7.93	99.5	62.32	56.71	46.92
SnapKV2026.04		31.15	44.11	54.55	55.47	45.14	31.14	31.22	24.94	24.09	70.5	91.9	42.23	7.96	99.5	63.55	57.91	48.46
FastKV2026.04		30.54	40.75	54.57	54.33	46.3	30.55	28.15	24.29	21.91	73	92.38	42.96	7.36	99.5	60.11	54.79	47.59
FastKV2026.04		30.26	43.7	54.83	54.39	46.42	30.42	30.28	24.88	22.93	73.5	91.97	43.07	7.61	99.5	61.61	55.67	48.19
Full-context2026.04		30.21	45.53	55.01	56.01	46.65	31.28	35.13	25.28	27.25	73	91.64	43.8	8.91	99.5	63.38	56.64	49.33
StreamingLLM2026.04		27.99	28.89	34.42	51.21	41.03	24.43	28.34	22.08	22.16	64	89.96	42.27	8.66	85	62.02	56.85	43.08
StreamingLLM2026.04		26.05	26.29	33.68	48.07	37.93	25.03	25.37	21.49	19.97	58.5	88.46	42.61	7.91	88	61.07	54.94	41.59
GemFilter2026.04		26.02	30.74	47.3	37.97	55.97	20.64	31.19	20.92	21.06	61.5	92.75	40.92	6.18	97	31.99	41.12	41.42
GemFilter2026.04		24.36	21.07	39.73	25.99	43.92	35.78	28.94	21.42	17.62	61	91.53	40.88	4.76	85	59.95	44.38	40.4
PyramidInfer2026.04		12.1	35.54	37.58	17.66	33.71	6.56	27.84	21.62	21.04	64	91.89	43.31	3.65	70.08	64.92	57.9	38.09
H2O2026.04		9.8	41.09	41	19.2	33.92	6.83	32.26	21.5	25.62	66	91.27	42.41	4.31	68.25	62.02	54.08	38.72
H2O2026.04		9.32	34.96	37.08	19.65	32.65	6.91	31.29	21.01	25.59	66.5	91.03	42.16	3.41	65.75	58.27	51.48	37.32