Long-context language evaluation on LongBench v1 (test)

19.6NrtQA ScoreLOOKAHEADKV

Evaluation Results

Method
LOOKAHEADKV2026.03	19.6	24.3	45.69	38.81	34.02	17.91	25.51	23.11	23.15	70	85.37	42.16	0.12	91	45.29	38.98	39.06
PyramidKV2026.03	19.52	24.01	44.53	39.48	31.94	16.37	26.16	22.68	24.68	72	84.84	41.79	0.5	95	46.57	40.12	39.39
LOOKAHEADKV2026.03	19.46	23.27	44.59	37.81	33.82	17.97	23.71	23.12	21.7	65.5	85.56	42.38	0.12	92.75	44.8	38.61	38.45
SnapKV2026.03	19.44	24.89	45.18	38.97	32.65	17.32	27.89	22.6	24.81	72.5	85.21	42.37	0.17	95	46.5	40.43	39.75
SnapKV2026.03	19.32	22.22	44.07	36.25	30.04	15.66	22.2	22.15	21.73	70.5	84.81	40.54	0.14	94.5	46.75	40.26	38.2
LOOKAHEADKV2026.03	19.2	25.15	44.48	39.09	32.85	18.39	29.18	23.24	25.14	73	84.84	43.19	0.17	95	46.24	39.86	39.94
LOOKAHEADKV2026.03	19.12	21.73	44.25	37.18	33.01	15.98	21.42	22.53	19.73	56.5	85.56	41.72	0	89	44.58	38.05	36.9
LOOKAHEADKV2026.03	19.08	25.15	45.04	38.79	33	17.85	29.86	23	25.26	73.5	85.21	43.06	0.17	94	46.62	40.53	40.01
LOOKAHEADKV2026.03	19.04	24.66	44.68	39.04	33.66	17.64	27.46	23.31	24.17	73	85.37	42.87	0.17	94	45.27	39.25	39.6
FullKV2026.03	18.94	24.78	46.15	39.15	33.11	19.03	30.33	23.06	25.18	74	85.21	42.87	0	94.5	46.82	40.24	40.21
PyramidKV2026.03	18.76	23.44	43.47	36.96	29.98	16.21	23.08	22.21	22.68	72.5	84.76	40.19	0.5	95.5	46.89	39.98	38.57
LAQ2026.03	18.74	21.71	45.85	37.93	33.55	16.16	22.63	23.05	21.51	70	85.21	42.73	0.17	95	41.83	41.63	38.61
SnapKV2026.03	18.68	24.05	44.25	38.57	30.72	16.63	24.97	22.28	23.62	71.5	85.26	40.43	0.5	95	46.39	39.03	38.87
LAQ2026.03	18.38	21.08	45.04	38.04	33.52	15.36	20.17	22.7	18.78	62.5	85.21	41.79	0.14	92	42.94	40.84	37.41
PyramidKV2026.03	17.95	20.69	41.43	36.22	29.68	14.96	20.39	21.52	20.01	66	84.65	40.16	0.14	93.5	45.61	38.83	36.98
SnapKV2026.03	17.52	20.72	39.73	34.12	30.25	14.94	19.06	21.99	19.43	61	83.82	38.53	0	94.5	44.98	39.87	36.28
PyramidKV2026.03	17.07	19.74	38.19	33.18	29.09	13.95	17.92	21.26	17.67	56	81.47	39.26	0	91.5	45.19	38.23	34.98
LAQ2026.03	16.99	22.67	46.97	38.1	33.62	16.38	24.58	23.39	22.99	71.5	84.71	42.25	0	94	40.85	40.3	38.71
LAQ2026.03	16.76	22.6	45.45	38.65	33.44	17.15	26.68	23.32	24.11	73	84.71	43.11	0	95	39.42	41.02	39.03
LAQ2026.03	16.64	16.99	43.66	34.19	31.41	14	16.09	20.94	13.76	48	83.63	38.89	0	70.75	42.51	39.08	33.16
LAQ2026.03	16.42	22.65	45.81	38.66	33.44	17.38	28.79	23.41	25.14	73	84.71	42.94	0	95	43.38	40.3	39.44
StreamingLLM2026.03	16.32	22.12	29.96	31.34	28.13	9.37	26.32	20.96	24.58	68.5	84.59	42.84	0	31	46.63	39.92	32.66
SnapKV2026.03	15.94	18.73	38.02	33.35	28.02	11.74	15.4	20.8	15.71	47	81.65	38.77	0	86	45.47	38.55	33.45
StreamingLLM2026.03	15.32	18.63	26.9	27.88	26.44	8.47	23.56	20.34	23.82	65.5	84	41.69	0	18	46.93	39.71	30.45
PyramidKV2026.03	14.13	18.14	35.49	30.97	26.99	10.53	14.27	20.5	14.29	45.5	78.72	38.06	0	78	44.05	37.95	31.72
StreamingLLM2026.03	13.92	18.4	25.1	24.91	24.77	7.67	20.38	19.63	20.78	61	81.8	40.38	0	12.5	47.28	38.62	28.57
StreamingLLM2026.03	13.58	16.57	22.67	22.16	24.16	7.54	13.01	19.26	14.8	44.5	70.96	37.72	0	12.5	46.07	38.3	25.24
StreamingLLM2026.03	13.41	17.66	22.58	23.72	23.88	7.65	16.15	19.24	17.97	47.5	76.22	40.17	0	13.5	46.27	37.64	26.47
PyramidKV2026.03	13.2	16.31	27.19	23.83	25.8	6.95	11.23	19.12	10.72	37	73.99	35.99	0	43	43.01	35.97	26.46
SnapKV2026.03	13.01	17.5	30.47	28.14	25.68	8.63	11.89	19.77	11.03	37.5	78.59	37.62	0	54.5	44.42	36.92	28.48
StreamingLLM2026.03	12.25	16.83	20.38	22.62	24.61	7.26	11.43	18.96	12.12	39	66.48	35.14	0	12.5	42.95	36.9	23.71