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

Long-context understanding on RULER (14-Metric Evaluation)

100S1 Score

Full KV

Updated 24d ago

Evaluation Results

Method	Links
Full KV 2026.03		100	100	100	99	99	99	98.75	98.5	99.2	17.8	87	87	54	87.63
Palu 2026.03		100	98	75	97	78	30	73	64.25	86	15.3	78.33	53	38	68.14
EchoKV 2026.03		100	100	100	99	99	94	89.5	99.5	87.2	0.6	80	85	52	83.52
EchoKV-Hybrid 2026.03		100	100	100	99	99	99	97	100	97.4	13.8	79.67	85	54	86.45
Full KV 2026.03		100	92	100	84	80	57	92.25	92.25	95.8	80.7	89.67	65	47	82.74
Palu 2026.03		100	97	98	76	51	26	88.5	80.75	83	32.1	85	41	33	68.56
EchoKV 2026.03		100	88	99	82	59	19	62.75	67.5	85.2	46.6	81.67	65	43	69.13
EchoKV-Hybrid 2026.03		100	89	100	83	77	42	92.5	90.25	90.8	66.3	93.67	67	47	79.89
EchoKV-Hybrid 2026.03		99	80	88	76	53	10	48	50.75	80	27.1	82.33	61	45	61.55
CommonKV 2026.03		92	94	83	97	67	1	87	94.25	79.8	0.1	85.67	75	50	69.68
CommonKV 2026.03		86	46	36	27	18	0	7	12.75	66.6	15.1	75	46	44	36.88
EchoKV 2026.03		83	81	72	70	79	9	57.25	88	37	0	75.67	82	50	60.3
EchoKV 2026.03		77	64	28	54	14	1	6	9	39.8	15.3	51	56	41	35.08
EchoKV-Hybrid 2026.03		57	79	46	88	78	17	71.25	93.5	31.4	0.1	67.67	82	50	58.53
CommonKV 2026.03		45	28	0	13	0	0	7.5	8.5	38.4	9.9	39.33	21	30	18.51
Palu 2026.03		40	27	0	4	0	0	1.75	1	8.4	4.3	37.67	12	24	12.32
ThinK* 2026.03		21	58	2	32	17	5	20.25	14	44.2	17.1	89.67	60	42	32.47
Palu 2026.03		11	0	0	0	0	0	0	0	0	0.3	1	0	0	0.95
CommonKV 2026.03		11	0	0	2	0	0	0	1.25	1.6	12.3	27	15	25	7.32
ThinK* 2026.03		1	2	0	2	0	0	1.75	3.25	0.6	0	48	63	40	12.43