Multi-Hop Question Answering on HotpotQA (Noisy)
31.5PrecisionCraniMem
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CraniMemBackbone=Qwen2.5-7B-Instruct2026.03 | 31.5 | 31.7 | 31.2 | 112.44 | 1.1 | |
| Mem0Backbone=Qwen2.5-7B-Instruct2026.03 | 20.5 | 28.1 | 19.8 | 10.387 | 3.6 | |
| Vanilla RAGBackbone=Qwen2.5-7B-Instruct2026.03 | 5.8 | 33.9 | 6.8 | 58.968 | 2.7 |