Knowledge-based Reasoning on Filtered CL-Bench (test)
34.96Accuracy (DK)No Memory
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| No MemoryModel=Qwen3-4B2026.06 | 34.96 | 10 | 25.71 | 36.36 | 30.45 | |
| No MemoryModel=Qwen3-8B2026.06 | 34.96 | 6.67 | 35.71 | 34.85 | 32.18 | |
| TMEMModel=Qwen3-4B2026.06 | 34.96 | 16.67 | 34.29 | 34.85 | 32.87 | |
| TMEMModel=Qwen3-8B2026.06 | 34.15 | 10 | 42.86 | 34.85 | 33.91 | |
| MemAgentModel=Qwen3-8B2026.06 | 33.33 | 6.67 | 38.57 | 34.85 | 32.18 | |
| AMEMModel=Qwen3-8B2026.06 | 31.71 | 10 | 30 | 30.3 | 28.72 | |
| AMEMModel=Qwen3-4B2026.06 | 28.46 | 10 | 25.71 | 33.33 | 26.99 | |
| MemAgentModel=Qwen3-4B2026.06 | 26.02 | 10 | 30 | 33.33 | 26.99 |