Memory Retrieval on LongMemEval n=500 (test) benchmark leaderboard · SOTA2 Research

Memory Retrieval on LongMemEval n=500 (test)

80.43MAPMemReranker-4B

Evaluation Results

Method	Links
MemReranker-4B2026.05		80.43	83.15	78	80.69	83.54	83.69	82.03	87.34	89.6	70.39
MemReranker-0.6B2026.05		75.38	78.76	70.2	75.62	79.87	80.23	78.72	85.08	89.69	63.75
Gemini-3-Flash2026.05		72.59	75.75	66.6	72.22	77.33	78.08	75.56	82.8	89.77	60.1
BGE-v2-m32026.05		70.69	75.1	66	70.51	75.73	76.96	74.12	80.95	88.74	56.34
Qwen3-Reranker-8B2026.05		69.66	73.4	63	68.51	75.51	76.17	72.1	81.08	89.31	54.9
Qwen3-Reranker-0.6B2026.05		69.55	73.67	62.6	69	75.2	76.16	73.13	81.35	89	55.67
Qwen3-Reranker-4B2026.05		65.01	69.37	56.4	64.03	71.85	72.9	68.72	79	89.34	49.81
GPT-4o-mini2026.05		56.84	63.85	53.22	55.51	62.65	66.76	57.81	64.63	83.49	44.44