Multi-Agent System Performance on Database
71.72Task Success Rate (TS)MARBLE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MARBLEModel=Qwen3-32B-Instruct2026.03 | 71.72 | 69.15 | 69.68 | |
| A-MemModel=Qwen3-next-80B2026.03 | 71.59 | 70.39 | 71.38 | |
| A-MemModel=Qwen3-32B-Instruct2026.03 | 71.43 | 73.23 | 71.2 | |
| W.o. MemoryModel=Deepseek-v3.22026.03 | 71 | 82.4 | 75.55 | |
| LLMA-MemModel=Deepseek-v3.22026.03 | 71 | 81.07 | 76.2 | |
| W.o. MemoryModel=Qwen3-32B-Instruct2026.03 | 71 | 66.07 | 67.87 | |
| LLMA-MemModel=Qwen3-next-80B2026.03 | 70.41 | 68.55 | 69.98 | |
| LLMA-MemModel=Qwen3-32B-Instruct2026.03 | 70.41 | 76.18 | 71.06 | |
| LLMA-MemModel=Claude-Sonnet-4.52026.03 | 70 | 88.27 | 78.52 | |
| MARBLEModel=Deepseek-v3.22026.03 | 70 | 75.6 | 72.62 | |
| MARBLEModel=Claude-Sonnet-4.52026.03 | 69.7 | 82.96 | 75.7 | |
| A-MemModel=Claude-Sonnet-4.52026.03 | 69.7 | 82.31 | 75.13 | |
| W.o. MemoryModel=Claude-Sonnet-4.52026.03 | 69.39 | 88.03 | 77.35 | |
| W.o. MemoryModel=Qwen3-next-80B2026.03 | 69 | 72.32 | 67.61 | |
| A-MemModel=Deepseek-v3.22026.03 | 68.69 | 73.48 | 70.51 | |
| MARBLEModel=Qwen3-next-80B2026.03 | 67 | 68.53 | 64.38 |