Long-context understanding on LongBench v0.2

26.95Na.QA (NER QA)SpindleKV

Evaluation results

Method
SpindleKV2025.07	26.95	31.23	49.01	41.89	26.9	18.6	29.92	24.42	24.56	71.5	85.98	43.39	2.74	86.18	56.39	54.13	42.14
SpindleKV2025.07	26.69	30.19	49.32	42.09	27.23	18.69	28.52	24.19	23.61	71	86.38	43.54	3.03	87.18	55.19	53.62	41.93
PyramidKV2025.07	26.31	28.64	49.12	41.66	25.98	19.02	26.38	23.91	22.66	70	85.88	42.53	2.69	86.32	54.04	53.36	41.16
PyramidKV2025.07	26.12	27.31	48.31	41.44	25.14	18.72	25.28	23.61	21.99	69	86.27	42.65	2.53	84.81	53.71	52.77	40.6
SpindleKV2025.07	26.03	29.49	49.61	41.87	26.5	17.98	27.01	23.73	22.73	70	86.28	43.59	2.29	86.99	53.81	53.57	41.34
SpindleKV2025.07	26.02	28.05	48.84	40.74	25.45	19.05	25.51	23.72	21.95	69.5	86.13	42.47	2.7	85.65	53.89	52.48	40.76
SpindleKV2025.07	25.97	30.34	49.17	42.06	27.35	18.22	27.99	24.25	23.19	70	86.22	42.96	2.74	87.26	54.69	53.77	41.64
FullKV2025.07	25.7	29.75	41.12	45.55	35.87	22.35	25.63	23.03	26.21	73	90.56	41.88	4.67	69.25	58.05	50.77	41.46
PyramidKV2025.07	25.14	26.11	46.97	40.56	25.46	19.06	25	23.32	21.55	70.5	86.41	41.92	3.26	84.56	51.95	52.23	40.25
PyramidKV2025.07	24.96	25.19	47.12	39.94	25.45	18.63	24.05	23.35	20.71	69.5	85.48	41.87	2.48	83.56	52.02	51.35	39.73
PyramidKV2025.07	24.36	24.66	45.85	40.67	24.81	17.83	23.29	23.41	20.53	70	86.09	40.74	3.27	81.84	51.54	50.6	39.34
PyramidInfer2025.07	23.75	17.46	29.97	35.08	23.92	16.9	28.08	21.26	24.42	62	85.06	41.45	1.04	41.23	50.95	52.86	34.71
PyramidInfer2025.07	22.65	14.57	30.14	33.87	23.52	15.59	26.82	21.1	24.89	61	84.18	40.57	1.85	32.25	50.64	53.02	33.43
PyramidInfer2025.07	21.43	13.49	25.88	31.92	20.44	14.83	25.27	20.57	24.02	58	82.16	40.77	1.45	27.93	52.53	52.55	32
PyramidInfer2025.07	18.73	11.96	24.34	27.1	15.87	11.98	24.89	20.07	21.19	54	76.19	39.92	2.06	20.83	51.83	52.5	29.59
PyramidInfer2025.07	17.28	11.52	23.41	26.38	16.72	11.98	24.49	19.41	20.77	53	68.87	39.78	3.53	11.88	54.48	52.72	28.51