Question Answering on MuSiQue (LM, F1, Acc., Rounds, Time)
29.01LM ScoreGoldRouter
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GoldRouterModel=Qwen-Plus2025.01 | 29.01 | 33.75 | 47.55 | 1.51 | 29.87 | |
| RIRSModel=Mixed, Note=RAG agents use LLaMA-3.1-8B, other modules use Qwen-Plus2025.01 | 24.5 | 28.57 | 39.75 | 2.08 | 47.51 | |
| ChameleonModel=Qwen-Plus2025.01 | 24.21 | 28.88 | 39.87 | 1.77 | 48.46 | |
| RIRSModel=Qwen-Plus2025.01 | 23.81 | 28.63 | 41.24 | 1.67 | 43.98 | |
| RouterDCModel=Qwen-Plus2025.01 | 23.38 | 27.69 | 37.43 | 1.71 | 52.98 | |
| GoldRouterModel=LLaMA-3.1-8B2025.01 | 23.27 | 28.18 | 31.46 | 1.86 | 21.68 | |
| RIRSModel=LLaMA-3.1-8B2025.01 | 20.16 | 24.22 | 25.82 | 2.48 | 25.14 | |
| ChameleonModel=LLaMA-3.1-8B2025.01 | 15.69 | 20.11 | 23.74 | 2.08 | 32.81 | |
| RouterDCModel=LLaMA-3.1-8B2025.01 | 15.08 | 19.11 | 21.58 | 1.98 | 29.66 |