Long-context language understanding on LongBench (Full Metric Suite) benchmark leaderboard · SOTA2 Research

Long-context language understanding on LongBench (Full Metric Suite)

27.36Narrative Score (Nrtv.)PyramidKV

Updated 2mo ago

Evaluation Results

Method	Links
PyramidKV2026.04		27.36	40.8	49.26	30.64	46.99	33.73	28.34	31.24	35.3	25.62	27.76	15.46	65.63	88.33	34.35	33.59	6.25	86.72	75.78	55.6	55.14	42.57
SnapKV2026.04		25.05	31.87	47.28	28.34	52.36	37.21	24.86	27.83	35.93	25.88	27.64	15.87	64	92.47	38.98	50	4	98	92	53.28	49.38	43.92
SnapKV + ART2026.04		24.57	39.7	43.93	28.01	43.94	33.33	25.47	24.3	34.14	24.92	26.69	15.83	69.53	86.77	41.54	40.63	4.69	99.22	91.41	52.51	54.38	43.12
PyramidKV + ART2026.04		24.51	40.23	49.05	31.34	46.42	33.7	28.77	31.19	35.02	25.32	27.89	15.46	65.63	86.22	32.75	31.25	3.91	86.72	75	56.34	55.2	42
Baseline2026.04		24.34	41.59	52.18	33.43	46.34	33.97	26.32	31.96	35.41	25.54	27.74	15.54	77.34	89.08	47.47	39.84	3.91	98.44	94.53	63.8	63.33	46.29
ART2026.04		22.04	42.83	51.72	32.19	45.16	35.07	25.54	30.08	35.56	25.82	27.81	16.41	76.56	89.11	46.64	39.84	3.91	97.66	92.97	63.96	59.65	45.74
PyramidKV + ART2026.04		19.72	18.46	36.97	18.87	39.56	29.66	22.19	27.41	30.5	21.38	24.38	14.53	55.47	81.19	31.72	22.27	2.43	15.63	20.7	46.43	49.8	29.97
StreamingLLM2026.04		18.77	34.24	32.72	27.62	37.65	33.02	23.21	16.57	33.53	24.89	27.07	17.74	73.44	88.59	46.43	40.63	2.51	79.3	59.83	65.48	62.84	40.29
StreamingLLM2026.04		18.65	20.38	24.81	16.34	29.19	21.98	12.95	14.65	28.73	22.01	23.43	16.69	68.75	86.39	44.57	26.56	1.99	29.95	22.66	63.13	59.41	31.11
PyramidKV2026.04		18.62	19.38	36.63	18.59	41.41	29.03	23.13	27.68	30.51	21.37	24.4	14.2	54.69	83.54	31.21	20.7	2.53	15.63	19.92	46.95	51.41	30.07
StreamingLLM + ART2026.04		18.39	20.38	24.81	16.34	29.19	21.46	12.95	14.81	28.69	22.01	23.43	16.66	68.75	87.03	44.64	26.56	1.99	29.95	22.66	63.13	59.08	31.09
SnapKV2026.04		17.89	18.45	27.76	15.72	23.09	22.11	13.55	18.65	26.89	20.54	21.64	15.08	44.53	78.65	35.82	28.13	4.37	99.22	29.43	46.7	46.28	31.17
StreamingLLM + ART2026.04		17.72	34.32	32.78	27.39	36.17	32.82	24.22	16.13	33.88	24.87	27.07	17.5	73.44	88.98	46.65	40.63	2.61	79.3	60.61	65.41	61.64	40.2
SnapKV + ART2026.04		16.19	18.71	28.12	16.07	24.43	21.63	12.08	18.16	27.27	20.65	21.53	14.82	42.97	78.01	35.22	25.78	2.71	99.22	30.34	45.8	46.99	30.8