MemoryBench Task on MemoryBench Long-Input-Short-Output
52.6Norm-ScoreUNO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UNOModel=phi-4, Method Type=Training2026.02 | 52.6 | 18.4 | |
| UNOModel=Qwen3-8B, Method Type=Training2026.02 | 49.99 | 13.49 | |
| SFT w/ UNO DataModel=Qwen3-8B, Method Type=Training2026.02 | 48.8 | 7 | |
| EmbeddingModel=Qwen3-8B, Method Type=RAG2026.02 | 48.41 | 12.29 | |
| BM25Model=Qwen3-8B, Method Type=RAG2026.02 | 48.4 | 10.87 | |
| A-MemModel=Qwen3-8B, Method Type=Memory2026.02 | 47.94 | 5.91 | |
| UNO-SingleModel=phi-4, Method Type=Training2026.02 | 47.16 | 1.93 | |
| MemOSModel=Qwen3-8B, Method Type=Memory2026.02 | 47.14 | -12.16 | |
| DPO w/ UNO DataModel=Qwen3-8B, Method Type=Training2026.02 | 47.14 | 2.3 | |
| BaseModel=Qwen3-8B, Method Type=Baseline2026.02 | 46.94 | 6.02 | |
| UNO-SingleModel=Qwen3-8B, Method Type=Training2026.02 | 46.81 | 5.34 | |
| BaseModel=phi-4, Method Type=Baseline2026.02 | 46.46 | -1.59 | |
| ReMemModel=Qwen3-8B, Method Type=Memory2026.02 | 43.78 | -7.7 | |
| BM25Model=phi-4, Method Type=RAG2026.02 | 43.26 | -15.07 | |
| EmbeddingModel=phi-4, Method Type=RAG2026.02 | 42.91 | -9.62 | |
| MemOSModel=phi-4, Method Type=Memory2026.02 | 42.83 | -26.43 | |
| A-MemModel=phi-4, Method Type=Memory2026.02 | 39.69 | -34.26 | |
| SFT w/ UNO DataModel=phi-4, Method Type=Training2026.02 | 38.97 | -30.81 | |
| ReMemModel=phi-4, Method Type=Memory2026.02 | 38.78 | -35.81 | |
| MemoryOSModel=Qwen3-8B, Method Type=Memory2026.02 | 38.51 | -24.87 | |
| DPO w/ UNO DataModel=phi-4, Method Type=Training2026.02 | 32.59 | -54.2 | |
| MemoryOSModel=phi-4, Method Type=Memory2026.02 | 25.59 | -89.25 |