Single-hop Question Answering on LoCoMo Single-Hop
50.45F1 ScoreMEMMA_LM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MEMMA_LMBackbone=GPT-4o-mini2026.03 | 50.45 | 38.66 | 82.86 | — | |
| MEMMA_LMBackbone=Claude-Haiku-4.52026.03 | 45.31 | 35.85 | 77.14 | — | |
| LightMemBackbone=Claude-Haiku-4.52026.03 | 44.06 | 36.56 | 71.43 | — | |
| LightMemBackbone=GPT-4o-mini2026.03 | 43.88 | 34.68 | 78.57 | — | |
| A-MemBackbone=GPT-4o-mini2026.03 | 42.72 | 32.43 | 62.86 | — | |
| Full TextBackbone=GPT-4o-mini2026.03 | 41.45 | 29.96 | 74.29 | — | |
| Full TextBackbone=Claude-Haiku-4.52026.03 | 41.45 | 29.96 | 74.29 | — | |
| Naive RAGBackbone=GPT-4o-mini2026.03 | 39.32 | 30.35 | 58.57 | — | |
| Naive RAGBackbone=Claude-Haiku-4.52026.03 | 39.32 | 30.35 | 58.57 | — | |
| A-MemBackbone=Claude-Haiku-4.52026.03 | 38.37 | 27.98 | 65.71 | — | |
| AriadneMem2026.07 | 26.63 | — | 16.67 | 22.27 | |
| LangMemBackbone=GPT-4o-mini2026.03 | 23.52 | 17.59 | 35.71 | — | |
| LangMemBackbone=Claude-Haiku-4.52026.03 | 22.65 | 16.19 | 48.57 | — | |
| SeCom2026.07 | 13.25 | — | 17.73 | 10.39 | |
| A-MemA-TMA augmentation=true2026.07 | 11.91 | — | 43.14 | 12.14 | |
| A-MemA-TMA augmentation=false2026.07 | 11.76 | — | 43.14 | 10.04 | |
| InsideOutA-TMA augmentation=true2026.07 | 11.58 | — | 17.38 | 8.31 | |
| Graphiti/ZepA-TMA augmentation=true2026.07 | 10.6 | — | 34.4 | 8.69 | |
| Graphiti/ZepA-TMA augmentation=false2026.07 | 10.51 | — | 24.82 | 8.88 | |
| M+2026.07 | 10.09 | — | 12.41 | 8.12 | |
| MemoryLLM2026.07 | 9.73 | — | 12.06 | 7.53 | |
| AtomMem2026.07 | 9.72 | — | 14.54 | 3.32 | |
| InsideOutA-TMA augmentation=false2026.07 | 7.37 | — | 34.04 | 5.48 | |
| Mem02026.07 | 4.62 | — | 56.38 | 3.45 |