MMD Performance and Efficiency on MMD
485Generation ScoreGreedy
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Greedythreshold (tau)=0.95, Base LLM=Qwen3-8B2026.05 | 485 | 656.81 | |
| Greedythreshold (tau)=0.90, Base LLM=Qwen3-8B2026.05 | 296 | 436.55 | |
| Greedythreshold (tau)=0.80, Base LLM=Qwen3-8B2026.05 | 212 | 318.46 | |
| Randthreshold (tau)=0.95, Base LLM=Qwen3-8B2026.05 | 132 | 199.95 | |
| OpenEvolvethreshold (tau)=0.80, Base LLM=Qwen3-8B2026.05 | 119 | 181.48 | |
| Randthreshold (tau)=0.90, Base LLM=Qwen3-8B2026.05 | 119 | 179.07 | |
| EXP3.Pthreshold (tau)=0.95, Base LLM=Qwen3-8B2026.05 | 114 | 172.74 | |
| Thompsonthreshold (tau)=0.95, Base LLM=Qwen3-8B2026.05 | 106 | 146.75 | |
| Thompsonthreshold (tau)=0.90, Base LLM=Qwen3-8B2026.05 | 105 | 144.75 | |
| EXP3.Pthreshold (tau)=0.90, Base LLM=Qwen3-8B2026.05 | 95 | 142.59 | |
| UCBthreshold (tau)=0.90, Base LLM=Qwen3-8B2026.05 | 92 | 117.63 | |
| UCBthreshold (tau)=0.95, Base LLM=Qwen3-8B2026.05 | 92 | 117.63 | |
| ShinkaEvolvethreshold (tau)=0.80, Base LLM=Qwen3-8B2026.05 | 71 | 112 | |
| EXP3.Pthreshold (tau)=0.80, Base LLM=Qwen3-8B2026.05 | 40 | 52.72 | |
| Randthreshold (tau)=0.80, Base LLM=Qwen3-8B2026.05 | 26 | 32.34 | |
| UCBthreshold (tau)=0.80, Base LLM=Qwen3-8B2026.05 | 8 | 12.41 | |
| Thompsonthreshold (tau)=0.80, Base LLM=Qwen3-8B2026.05 | 8 | 12.41 |