Temporal Question Answering on LoCoMo Temporal
59.76F1LightMem
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LightMemBackbone=GPT-4o-mini2026.03 | 59.76 | 51.12 | 78.38 | |
| MEMMA_LMBackbone=Claude-Haiku-4.52026.03 | 59.25 | 44.66 | 83.78 | |
| LightMemBackbone=Claude-Haiku-4.52026.03 | 58.42 | 49.85 | 89.19 | |
| MEMMA_LMBackbone=GPT-4o-mini2026.03 | 57.21 | 41.94 | 83.78 | |
| A-MemBackbone=Claude-Haiku-4.52026.03 | 56.41 | 43.23 | 54.05 | |
| A-MemBackbone=GPT-4o-mini2026.03 | 55.01 | 42.4 | 51.35 | |
| LangMemBackbone=Claude-Haiku-4.52026.03 | 34.72 | 26.33 | 37.84 | |
| Full TextBackbone=GPT-4o-mini2026.03 | 29.95 | 19.33 | 51.35 | |
| Full TextBackbone=Claude-Haiku-4.52026.03 | 29.95 | 19.33 | 51.35 | |
| Naive RAGBackbone=GPT-4o-mini2026.03 | 17.3 | 12.36 | 35.14 | |
| Naive RAGBackbone=Claude-Haiku-4.52026.03 | 17.3 | 12.36 | 35.14 | |
| LangMemBackbone=GPT-4o-mini2026.03 | 15.23 | 11.53 | 21.62 |