Share your thoughts, 1 month free Claude Pro on usSee more

Long-context evaluation on LongBench (test)

32.85NarQA Score

xKV

Updated 16d ago

Evaluation Results

Method	Links
xKV 2025.03		32.85	45.62	54.69	51.74	36.38	28.33	31.32	24.49	22.61	26.26	100	41.3
PyramidKV 2025.03		31.83	44.36	52.56	56.38	48.35	27.23	28.63	25.06	23.49	24.8	99.5	42.02
TokenButler 2025.03		30.73	48.86	56.45	57.77	50.82	28.34	34.34	25.3	26.91	29.22	100	44.43
Dense 2025.03		30.64	46.64	55.42	58.17	47.76	28.23	34.51	25.61	25.43	23.51	100	43.27
SnapKV 2025.03		30.34	45.56	53.62	57.92	47.5	26.96	28.99	25.33	23.95	24.91	99.5	42.23
PyramidKV 2025.03		29.93	41.29	47.21	60.16	51.98	32.79	27.21	22.88	19.86	45	100	43.48
KIVI 2025.03		29.84	41.49	51.28	57.27	42.54	27.54	33.28	25.01	23.9	25.88	98.17	41.47
TokenButler 2025.03		29.39	47.27	52.5	59.67	56.01	34.58	33.32	23.14	24.04	35.09	100	45
Dense 2025.03		29.21	43.78	48.58	60.92	53.29	33.68	33.23	23.24	23.53	43.21	100	44.79
xKV 2025.03		28.78	42.98	47.43	58.79	52.42	32.69	33.02	23.67	23.18	38.62	98	43.6
SnapKV 2025.03		28.68	44.51	47.43	60.88	51.91	32.17	29.95	23.05	21.79	45.67	100	44.19
Single SVD 2025.03		27.88	41.84	47.66	52.73	48.76	28.46	28.71	22.24	21.18	33.18	65	37.97
KIVI 2025.03		26.43	35.78	38.4	46.17	45.08	23.79	24.62	22.41	16.98	39.52	63.5	34.79
StreamingLLM 2025.03		26.25	29.32	34.56	49.99	42.47	20.92	26.29	21.9	22.63	23.87	93.5	35.61
StreamingLLM 2025.03		24.13	34.99	27.47	47.55	45.13	21.44	26.76	19.34	21.15	44.11	29	31.01
MiniCache 2025.03		18.11	20.41	27.51	38.8	21.37	20.19	21.09	20.85	19.6	22.75	98	29.88
Single SVD 2025.03		15.52	38.36	37.14	30.04	19.7	22.06	19.52	22.35	18.91	22.82	85	30.13
MiniCache 2025.03		7.98	16.47	9.42	21.78	15.46	10.55	8.83	7.94	3.84	11.32	5	10.78