Person understanding and persistent memory on CUPID
63.5ScorePERSONATREE
Evaluation Results
| Method | Links | |
|---|---|---|
| PERSONATREEBackbone=GPT-5.4 Mini2026.06 | 63.5 | |
| PERSONATREEBackbone=Gemini 3 Flash2026.06 | 61.4 | |
| A-MEMBackbone=GPT-5.4 Mini2026.06 | 60.8 | |
| TiMemBackbone=GPT-5.4 Mini2026.06 | 60.6 | |
| Mem0Backbone=Gemini 3 Flash2026.06 | 59.5 | |
| Mem0Backbone=GPT-5.4 Mini2026.06 | 59.5 | |
| TiMemBackbone=Gemini 3 Flash2026.06 | 59.1 | |
| Flat retrievalBackbone=GPT-5.4 Mini2026.06 | 58.3 | |
| A-MEMBackbone=Gemini 3 Flash2026.06 | 57.4 | |
| Full historyBackbone=GPT-5.4 Mini2026.06 | 56.8 | |
| Flat retrievalBackbone=Gemini 3 Flash2026.06 | 56.7 | |
| PERSONATREEBackbone=Qwen3-32B2026.06 | 55.8 | |
| Full historyBackbone=Gemini 3 Flash2026.06 | 55.1 | |
| TiMemBackbone=Qwen3-32B2026.06 | 53.1 | |
| Mem0Backbone=Qwen3-32B2026.06 | 51.8 | |
| A-MEMBackbone=Qwen3-32B2026.06 | 51.7 | |
| Flat retrievalBackbone=Qwen3-32B2026.06 | 50.1 | |
| Full historyBackbone=Qwen3-32B2026.06 | 48.8 |