Long-context Memory Retrieval and Reasoning on WebDancer 128K
38.21F1 ScoreMemSifter
Evaluation Results
| Method | Links | |
|---|---|---|
| MemSifterBackbone LLM=DeepSeek V3.22026.03 | 38.21 | |
| GemmaEmbBackbone LLM=DeepSeek V3.22026.03 | 37.82 | |
| RearankBackbone LLM=DeepSeek V3.22026.03 | 37.23 | |
| A-MEMBackbone LLM=DeepSeek V3.22026.03 | 36.72 | |
| NemoriBackbone LLM=DeepSeek V3.22026.03 | 35.28 | |
| MemSifterBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 35.1 | |
| DS-V3.2Backbone LLM=DeepSeek V3.22026.03 | 34.31 | |
| BGE-M3Backbone LLM=DeepSeek V3.22026.03 | 33.89 | |
| A-MEMBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 33.67 | |
| Qwen3-30BBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 33.63 | |
| HippoRAGBackbone LLM=DeepSeek V3.22026.03 | 33.26 | |
| ReasonRankBackbone LLM=DeepSeek V3.22026.03 | 32.95 | |
| Mem0Backbone LLM=Qwen3-30B-A3B-Ins2026.03 | 32.88 | |
| NemoriBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 32.77 | |
| Mem0Backbone LLM=DeepSeek V3.22026.03 | 32.58 | |
| GemmaEmbBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 32.45 | |
| BGE-M3Backbone LLM=Qwen3-30B-A3B-Ins2026.03 | 31.85 | |
| HippoRAGBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 31.79 | |
| RearankBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 31.4 | |
| ReasonRankBackbone LLM=Qwen3-30B-A3B-Ins2026.03 | 30 |