Long-context Memory on LOCOMO
48.88F1 ScoreTreeMem
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TreeMemBackbone=Qwen2.5-7B, Context window=32K2026.05 | 48.88 | 43.38 | |
| GAMBackbone=Qwen2.5-7B, Context window=32K2026.05 | 45.77 | 39.42 | |
| MemOSBackbone=GPT-4o-mini2026.04 | 45.56 | 38.06 | |
| OursBackbone=GPT-4o-mini2026.04 | 45.2 | 38.52 | |
| OursBackbone=Qwen2.5-72B2026.04 | 43.87 | 37.3 | |
| Memory-R1Backbone=Qwen2.5-7B, Context window=32K2026.05 | 43.14 | 36.14 | |
| ZepBackbone=GPT-4o-mini2026.04 | 42.88 | 36.09 | |
| MemOSBackbone=Qwen2.5-72B2026.04 | 42.79 | 35.16 | |
| MemoryOSBackbone=GPT-4o-mini2026.04 | 42.57 | 34.75 | |
| MemoryOSBackbone=Qwen2.5-72B2026.04 | 39.63 | 32.62 | |
| OursBackbone=Qwen2.5-7B2026.04 | 38.03 | 31.73 | |
| ZepBackbone=Qwen2.5-72B2026.04 | 37.45 | 30.46 | |
| MemOSBackbone=Qwen2.5-7B2026.04 | 37.05 | 30.3 | |
| OursBackbone=LLaMA3.1-8B2026.04 | 35.21 | 28.4 | |
| MemTreeBackbone=Qwen2.5-72B2026.04 | 34.85 | 28.49 | |
| ZepBackbone=Qwen2.5-7B2026.04 | 33.82 | 27.42 | |
| MemOSBackbone=LLaMA3.1-8B2026.04 | 33.39 | 26.24 | |
| MemTreeBackbone=GPT-4o-mini2026.04 | 32.32 | 25.89 | |
| MemoryOSBackbone=LLaMA3.1-8B2026.04 | 32.21 | 25.56 | |
| Mem0Backbone=Qwen2.5-7B, Context window=32K2026.05 | 30.61 | 23.55 | |
| MemTreeBackbone=Qwen2.5-7B2026.04 | 29.68 | 23.95 | |
| MemoryOSBackbone=Qwen2.5-7B2026.04 | 28.34 | 23.11 | |
| A-MemBackbone=Qwen2.5-7B, Context window=32K2026.05 | 26.08 | 21.78 | |
| MemTreeBackbone=LLaMA3.1-8B2026.04 | 23.49 | 18.19 | |
| RAGBackbone=Qwen2.5-7B, Context window=32K2026.05 | 8.97 | 7.27 |