Ordinal Preference Alignment on GLOBALQA
0.8911FI Bor.APPA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| APPAModel=gemma-2-2b2026.04 | 0.8911 | 0.511 | 0.461 | 1 | |
| APPAModel=llama-3.2-3b2026.04 | 0.8819 | 0.536 | 0.487 | 0.969 | |
| PPO/Bor.Model=gemma-2-2b, Server Agg.=MIN2026.04 | 0.8725 | 0.475 | 0.42 | 1 | |
| PPO/Bor.Model=gemma-2-2b, Server Agg.=AVERAGE2026.04 | 0.8539 | 0.469 | 0.359 | 1 | |
| PPO/Bor.Model=llama-3.2-3b, Server Agg.=AVERAGE2026.04 | 0.853 | 0.526 | 0.417 | 0.969 | |
| BASEModel=llama-3.2-3b, Server Agg.=-2026.04 | 0.8463 | 0.363 | 0.351 | 0.771 | |
| PPO/Bor.Model=llama-3.2-3b, Server Agg.=MIN2026.04 | 0.8453 | 0.451 | 0.431 | 0.969 | |
| BASEModel=Qwen3-0.6B-Instruct, Server Agg.=-2026.04 | 0.8315 | 0.326 | 0.277 | 0.959 | |
| SFTModel=gemma-2-2b, Server Agg.=-2026.04 | 0.8313 | 0.339 | 0.329 | 1 | |
| BASEModel=gemma-2-2b, Server Agg.=-2026.04 | 0.8265 | 0.461 | 0.434 | 0.999 | |
| SFTModel=llama-3.2-3b, Server Agg.=-2026.04 | 0.8238 | 0.461 | 0.45 | 0.934 | |
| PPO/Bor.Model=Qwen3-0.6B-Instruct, Server Agg.=MIN2026.04 | 0.8214 | 0.422 | 0.409 | 1 | |
| APPAModel=Qwen3-0.6B-Instruct2026.04 | 0.8214 | 0.442 | 0.429 | 1 | |
| SFTModel=Qwen3-0.6B-Instruct, Server Agg.=-2026.04 | 0.8169 | 0.36 | 0.351 | 1 | |
| PPO/Bor.Model=Qwen3-0.6B-Instruct, Server Agg.=AVERAGE2026.04 | 0.7819 | 0.416 | 0.353 | 1 |