Multi-hop Question Answering on LoCoMo Multi-Hop
48.15F1 ScoreMEMMA_LM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MEMMA_LMBackbone=GPT-4o-mini2026.03 | 48.15 | 39.67 | 78.12 | |
| MEMMA_LMBackbone=Claude-Haiku-4.52026.03 | 35.38 | 32.48 | 65.62 | |
| LightMemBackbone=Claude-Haiku-4.52026.03 | 35.11 | 31.85 | 59.38 | |
| LightMemBackbone=GPT-4o-mini2026.03 | 33.74 | 29.33 | 65.62 | |
| Full TextBackbone=GPT-4o-mini2026.03 | 29.41 | 21.16 | 43.75 | |
| Full TextBackbone=Claude-Haiku-4.52026.03 | 29.41 | 21.16 | 43.75 | |
| LangMemBackbone=Claude-Haiku-4.52026.03 | 20.05 | 14.85 | 34.38 | |
| Naive RAGBackbone=GPT-4o-mini2026.03 | 15.84 | 9.5 | 31.25 | |
| Naive RAGBackbone=Claude-Haiku-4.52026.03 | 15.84 | 9.5 | 31.25 | |
| A-MemBackbone=Claude-Haiku-4.52026.03 | 15.79 | 10.32 | 28.13 | |
| A-MemBackbone=GPT-4o-mini2026.03 | 15.56 | 10.88 | 31.25 | |
| LangMemBackbone=GPT-4o-mini2026.03 | 12.55 | 9.22 | 25 |