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

Long-context memory management on MemoryAgentBench

83Single-Doc Recall

TRUSTMEM

Updated 1mo ago

Evaluation Results

Method	Links
TRUSTMEM 2026.06		83	64	55	85	79.2	41	84	84	15.9	65.7	-
Fine-Mem 2026.01		76	72	55	69	80	59	89	82	15.3	66.4	-
Mem-α 2026.01		74	68	52	71	71	41	73	70	12.9	59.2	-
Mem-α 2026.06		74	68	52	71	71	41	73	70	12.9	59.2	-
RAG-Top2 2026.01		69	45	58.1	69	65	21	70	75	6.5	53.2	-
RAG-Top2 2026.06		69	45	58.1	69	65	21	70	75	6.5	50.2	-
Mem-T 2026.06		56	60	43.3	75	48	27	43	79	14.1	49.5	-
MemAgent 2026.06		51	52	23	62	78	30	88	83	10.9	53.1	-
A-Mem 2026.01		42	36	42.7	68	72	40	17	78	10.3	45.1	-
AtomMem 2026.06		37	33	8.7	11	1	9	0	0	18.9	13.2	-
Long-Context 2026.01		28	27	29.2	64	74	34	86	77	12.5	48	-
Long-Context 2026.06		28	27	29.2	64	74	34	86	77	12.5	46.1	-
LightMem 2026.01		27	32	42	45	61	25	49	48	1	36.7	-
MEM1 2026.06		23	33	13	57	80	28	48	76	7	40.6	-
MemAgent 2026.01		7	16	5	37	26	21	25	37	4.3	19.8	-
MEM1 2026.01		7	18	9	18	0	0	9	0	2.9	7.1	-