Regional knowledge and conversational settings on CARE
51.26Average ScoreLRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| LRPOBackbone=Gemma3-4b-it2026.05 | 51.26 | |
| GRPOBackbone=Gemma3-4b-it2026.05 | 51.24 | |
| DPOBackbone=Gemma3-4b-it2026.05 | 50.47 | |
| MAPOBackbone=Gemma3-4b-it2026.05 | 50.22 | |
| VanillaBackbone=Gemma3-4b-it2026.05 | 49.85 | |
| LIDRBackbone=Gemma3-4b-it2026.05 | 49.23 | |
| MPOBackbone=Gemma3-4b-it2026.05 | 48.98 | |
| LRPOBackbone=Qwen2.5-1.5b-it2026.05 | 34.6 | |
| GRPOBackbone=Qwen2.5-1.5b-it2026.05 | 33.66 | |
| DPOBackbone=Qwen2.5-1.5b-it2026.05 | 32.69 | |
| VanillaBackbone=Qwen2.5-1.5b-it2026.05 | 32.2 | |
| MAPOBackbone=Qwen2.5-1.5b-it2026.05 | 31.98 | |
| MPOBackbone=Qwen2.5-1.5b-it2026.05 | 31.91 | |
| LIDRBackbone=Qwen2.5-1.5b-it2026.05 | 30.07 | |
| GRPOBackbone=LLaMA3.2-1b-it2026.05 | 20.15 | |
| LRPOBackbone=LLaMA3.2-1b-it2026.05 | 18.62 | |
| MPOBackbone=LLaMA3.2-1b-it2026.05 | 17.67 | |
| VanillaBackbone=LLaMA3.2-1b-it2026.05 | 17.51 | |
| DPOBackbone=LLaMA3.2-1b-it2026.05 | 16.95 | |
| MAPOBackbone=LLaMA3.2-1b-it2026.05 | 16.84 | |
| LIDRBackbone=LLaMA3.2-1b-it2026.05 | 16.31 |