Long-term Memory Evaluation on LongMemEval single run
46.7F1 ScoreMemory-R1-GRPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Memory-R1-GRPOInference Run Type=single run2026.04 | 46.7 | 41.1 | 57.8 | |
| TSUBASA-PROMemory Evaluation Setting=static acc., Inference Run Type=single run, Number of retrieved items (k)=10, System Prompt Type=concise generation2026.04 | 45.75 | 42.1 | 57.4 | |
| TSUBASA-PROMemory Evaluation Setting=evolving mem., Inference Run Type=single run, Number of retrieved items (k)=10, System Prompt Type=concise generation2026.04 | 43.43 | 39.85 | 53 | |
| A-MemInference Run Type=single run2026.04 | 41.55 | 36.58 | 54.8 | |
| Memory-R1-PPOInference Run Type=single run2026.04 | 40.3 | 35.5 | 47.4 | |
| Mem0Inference Run Type=single run2026.04 | 38.44 | 34.53 | 46.8 | |
| LoCoMoInference Run Type=single run2026.04 | 18.27 | 14.57 | 22.2 |