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

30.04NrtvQA ScoreK-VEC

Evaluation results

Method
K-VEC2026.06	30.04	45.03	56.64	58.2	48.09	31.39	28.98	24.99	26.04	71.49	91.72	42.69	8.45	99.5	—	—	47.38
Pyramid2026.06	29.62	43.66	54.1	55.06	44.22	31.3	27.27	24.3	25.68	68.5	91.27	41.96	7.73	99.5	—	—	46.01
SnapKV2026.06	29.28	43.64	54.34	54.24	44.34	31.52	27.8	24.39	25.95	69	91.72	42.5	7.8	99.5	—	—	46.14
Ada-SnapKV2026.06	29.23	44.09	53.82	54.8	44.01	31.4	28.86	24.73	26.04	72.5	91.72	42.56	7.82	99.5	—	—	46.51
Ada-Pyramid2026.06	28.76	44.57	53.73	54.89	44.15	31.97	27.75	25.26	25.84	70.5	91.62	42.37	7.67	99.5	—	—	46.33
FullKV2026.05	27.96	44.61	49.84	58.84	43.27	25.54	30.64	22.56	24.06	74.5	87.15	45.36	2.15	100	64.69	57.56	47.42
SnapKV2026.05	27.8	42.45	48.56	59.01	43.04	25.83	26.55	23.26	22.91	72.5	87.44	42.8	1.83	100	64.83	56.65	46.59
RDKV2026.05	27.75	43.89	49.81	59.97	43.87	25.37	30.07	23.67	24	74.5	86.73	44.6	0.5	100	63.29	54.42	47.03
AdaKV2026.05	27.64	42.54	48.65	59.57	43.31	26.35	26.75	23.02	23.16	74	88.08	43.49	1.6	100	64.86	57.44	46.9
RDKV2026.05	27.35	44.36	50.4	59.35	43.63	24.87	30.91	23.29	24.45	74.5	87.26	45.24	0.75	100	62.86	54.61	47.12
RDKV2026.05	26.79	41.89	49.87	59.14	42.95	26.02	24.01	23.39	22.46	74	86.53	43	1.25	100	63.1	53.48	46.12
ThinK2026.05	26.73	42.98	47.88	58.62	43.18	24.97	27.9	23.55	23.43	75	87.85	42.95	1.97	100	65.48	57.04	46.85
AdaKV2026.05	26.72	40.87	48.17	59.1	42.93	26.61	23.28	22.58	21.54	70.5	87.15	42.2	1.78	100	64.69	56.58	45.92
Snap+Zip2026.05	26.6	43.46	47.8	59.57	43.76	22.11	29.78	22.67	23.84	73.5	87.36	44.65	2.16	99.5	64.09	56.08	46.68
SnapKV2026.05	26.57	39.86	48.52	59.27	43.23	26.1	23	22.74	21.67	70	87.95	41.81	1.99	100	64.72	56.35	45.86
ThinK2026.05	26.48	41.08	46.44	59.15	42.9	25.06	24.79	23.38	22.38	72.5	87.64	42.51	2.06	100	65.07	56.34	46.11
RDKV2026.05	26.45	43.49	50.3	59.6	42.91	25.52	27.3	23.8	23.3	74	87.03	44.27	1.25	100	63.39	53.91	46.66
SnapKV2026.05	25.92	36.55	48.56	59.14	41.07	25.18	19.13	22.75	19.34	65	84.21	40.26	1.72	99.75	62.46	53.74	44.05
Snap+Zip2026.05	25.47	42.27	47.8	57.58	43.17	23.24	28.44	22.6	23.79	75	86.82	43.79	1.94	99.5	65.03	56.52	46.44
AdaKV2026.05	25.46	38.2	48.1	59.34	42.49	26.44	19.46	23.01	19.6	69	86.32	40.8	1.75	100	63.18	55.07	44.89
ThinK2026.05	25.3	39.37	46.06	59.1	42.73	25.1	21.58	23.14	20.8	68.5	86.58	41.22	1.88	100	63.5	55.32	45.01
SLM2026.06	24.97	30.22	37.06	46.57	39.14	25.24	26.01	21.08	25.72	63.5	88.87	42.28	6.98	89	—	—	40.47
RDKV2026.05	24.94	39.23	49.24	59.71	43.46	25.5	20.01	23.15	20.18	70.5	84.64	41.88	0.75	100	60.43	52.29	44.74
K-VEC2026.06	24.55	40.02	47.23	55.98	41.49	29.22	28.63	23.79	24.78	55.73	85.99	42.57	3.01	84.91	12.19	10.95	38.19
Snap+Zip2026.05	24.21	40.99	48.26	58.93	42.75	24.22	25.96	22.62	23.13	72.5	87.13	43.42	1.9	99.5	63.77	54.68	45.87
SnapKV2026.06	23.88	38.59	46.43	55.12	42.3	27.57	27.09	23.1	24.56	53.45	84.87	39.82	2.11	84.5	10.43	8.67	37.03
Snap+Zip2026.05	23.84	38.35	48.62	57.59	41.88	24.04	23.43	22.18	21.62	68.5	85.73	42.24	1.91	98	62.43	53.41	44.61
AdaKV2026.05	23.78	31.71	45.03	57.81	40.54	23.84	16.4	22.7	16.45	56.5	81.1	40.47	1.6	98	60.64	51.76	41.77
K-VEC2026.06	23.45	34.15	39.45	51.83	40.34	26.54	22.28	23.38	18.9	49.23	85.44	41.23	2.89	84.5	11.83	10.23	35.35
ThinK2026.05	22.88	35.83	44.32	56.7	40.32	23.61	18.68	22.77	17.87	59	84.64	40.93	1.32	97	61.76	52.8	42.53
SnapKV2026.05	21.46	31.15	43.96	53.83	40.79	22.12	15.77	21.9	16.16	51.5	83.44	39.72	1.3	93.03	59.07	50.75	40.37
GemFilter2026.06	20.71	11	29.28	19.12	17.01	13.01	30.37	21.75	25.17	63	90.7	42.5	7.15	92.22	—	—	34.5
SnapKV2026.06	20.42	30.71	40.36	50.46	38.47	23.71	20.69	20.68	17.88	50.11	83.45	38.12	2.11	85	10.78	9.38	33.89
ThinK2026.05	18.65	28.98	39.41	51.5	39.33	19.25	14.72	20.81	14.53	44.5	80.45	38.43	0.93	85.5	55.67	48.79	37.59
Snap+Zip2026.05	18.22	31.79	46.14	52.48	38.27	20.22	20.43	21.16	19.33	61	84.02	40.83	1.67	60.29	60.95	51.23	39.25
SnapKV2026.05	16.95	26.1	35.04	43.94	34.68	15.7	12.69	19.63	12.69	39.5	74.68	35.99	1.51	45.5	53.03	45.78	32.09
AdaKV2026.05	16.83	25.08	36.52	45.79	35.94	15.06	12.97	19.4	12.97	42.5	76.67	36.46	2.73	47	53.54	45.54	32.81