Multi-turn dialogue on ShareGPT
94.11Success Rate (SR)MCTS Router
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MCTS RouterCandidate LLM Set=Qwen Series Set2026.04 | 94.11 | 4.97 | |
| MCTS RouterCandidate LLM Set=Llama Series Set2026.04 | 91.94 | 5.76 | |
| Greedy RouterCandidate LLM Set=Qwen Series Set2026.04 | 88.5 | 5.12 | |
| DialRouterCandidate LLM Set=Qwen Series Set2026.04 | 87.46 | 4.96 | |
| DialRouterCandidate LLM Set=Llama Series Set2026.04 | 85.82 | 5.72 | |
| Greedy RouterCandidate LLM Set=Llama Series Set2026.04 | 85.75 | 5.73 | |
| AvengersCandidate LLM Set=Qwen Series Set2026.04 | 83.93 | 5 | |
| KNN RouterCandidate LLM Set=Qwen Series Set2026.04 | 83.37 | 5.01 | |
| Qwen3-14BCandidate LLM Set=Qwen Series Set2026.04 | 83.24 | 5.11 | |
| Prompt LLMCandidate LLM Set=Qwen Series Set2026.04 | 82.85 | 4.92 | |
| Qwen3-8BCandidate LLM Set=Qwen Series Set2026.04 | 82.8 | 5.03 | |
| RouterDCCandidate LLM Set=Qwen Series Set2026.04 | 82.72 | 4.94 | |
| MF RouterCandidate LLM Set=Qwen Series Set2026.04 | 82.41 | 5.13 | |
| Qwen3-4BCandidate LLM Set=Qwen Series Set2026.04 | 81.75 | 5.09 | |
| KNN RouterCandidate LLM Set=Llama Series Set2026.04 | 81.71 | 5.77 | |
| RandomCandidate LLM Set=Qwen Series Set2026.04 | 81.5 | 4.95 | |
| RouterDCCandidate LLM Set=Llama Series Set2026.04 | 80.94 | 5.68 | |
| AvengersCandidate LLM Set=Llama Series Set2026.04 | 80.38 | 5.61 | |
| MF RouterCandidate LLM Set=Llama Series Set2026.04 | 80.01 | 5.67 | |
| Llama3.1-8B-InstructCandidate LLM Set=Llama Series Set2026.04 | 79.24 | 5.6 | |
| Prompt LLMCandidate LLM Set=Llama Series Set2026.04 | 79.19 | 5.79 | |
| RandomCandidate LLM Set=Llama Series Set2026.04 | 76.16 | 6.12 | |
| Llama3.2-3B-InstructCandidate LLM Set=Llama Series Set2026.04 | 73.77 | 5.92 | |
| Llama3.2-1B-InstructCandidate LLM Set=Llama Series Set2026.04 | 63.03 | 6.62 |