Long-horizon behavior prediction on OminiBehavior Ads
82.3AccuracyPraMem
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PraMemBackbone=GPT-OSS-120B, Category=Experiential Memory2026.07 | 82.3 | 43.7 | |
| PraMemBackbone=Qwen3.5-35B-A3B, Category=Experiential Memory2026.07 | 74.6 | 32.4 | |
| ProExBackbone=GPT-OSS-120B, Category=Advanced Memory2026.07 | 68.1 | 28.5 | |
| SummaryBackbone=GPT-OSS-120B, Category=Simple Memory2026.07 | 64.7 | 27.5 | |
| Mem0Backbone=GPT-OSS-120B, Category=Advanced Memory2026.07 | 63.1 | 29.5 | |
| MemOSBackbone=GPT-OSS-120B, Category=Advanced Memory2026.07 | 61.8 | 29 | |
| RAGBackbone=GPT-OSS-120B, Category=Simple Memory2026.07 | 61.3 | 22.2 | |
| ProExBackbone=Qwen3.5-35B-A3B, Category=Advanced Memory2026.07 | 61.3 | 24 | |
| TruncationBackbone=GPT-OSS-120B, Category=No Memory2026.07 | 61 | 26.2 | |
| Long-contextBackbone=GPT-OSS-120B, Category=No Memory2026.07 | 59.4 | 28.3 | |
| SummaryBackbone=Qwen3.5-35B-A3B, Category=Simple Memory2026.07 | 59.4 | 25.7 | |
| TruncationBackbone=Qwen3.5-35B-A3B, Category=No Memory2026.07 | 58 | 24.4 | |
| RAGBackbone=Qwen3.5-35B-A3B, Category=Simple Memory2026.07 | 55.9 | 25.1 | |
| Long-contextBackbone=Qwen3.5-35B-A3B, Category=No Memory2026.07 | 55.7 | 25.4 | |
| Mem0Backbone=Qwen3.5-35B-A3B, Category=Advanced Memory2026.07 | 55.6 | 24 | |
| MemOSBackbone=Qwen3.5-35B-A3B, Category=Advanced Memory2026.07 | 54.2 | 24.6 |