Long-context Memory Retrieval and Reasoning on LongMemEval 1M
49.58F1 ScoreMemSifter
Evaluation Results
| Method | Links | |
|---|---|---|
| MemSifterBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 49.58 | |
| BGE-M3Backbone LLM=Qwen3-30B-A3B-Ins2026.03 | 49.07 | |
| Mem0Backbone LLM=Qwen3-30B-A3B-Ins2026.03 | 48.53 | |
| ReasonRankBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 48.2 | |
| GemmaEmbBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 48.16 | |
| HippoRAGBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 47.62 | |
| NemoriBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 47.31 | |
| A-MEMBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 46.77 | |
| Qwen3-30BBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 46.42 | |
| RearankBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 45.6 | |
| MemSifterBackbone LLM=DeepSeek V3.22026.03 | 33.32 | |
| A-MEMBackbone LLM=DeepSeek V3.22026.03 | 32.49 | |
| RearankBackbone LLM=DeepSeek V3.22026.03 | 32.47 | |
| DS-V3.2Backbone LLM=DeepSeek V3.22026.03 | 30.89 | |
| GemmaEmbBackbone LLM=DeepSeek V3.22026.03 | 30.83 | |
| HippoRAGBackbone LLM=DeepSeek V3.22026.03 | 30.83 | |
| BGE-M3Backbone LLM=DeepSeek V3.22026.03 | 29.91 | |
| NemoriBackbone LLM=DeepSeek V3.22026.03 | 29.24 | |
| ReasonRankBackbone LLM=DeepSeek V3.22026.03 | 28.93 | |
| Mem0Backbone LLM=DeepSeek V3.22026.03 | 28.2 |