Personalized Response Generation on AIME Implicit Preference 2025 (test)
0.221Preference ScoreTriAlign
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TriAlign2026.06 | 0.221 | 0.511 | 0.467 | 0.025 | 0.044 | 0.049 | |
| SWIFTYear=20262026.06 | 0.219 | 0.408 | 0.333 | 0.059 | 0.075 | 0.145 | |
| SFTYear=20222026.06 | 0.217 | 0.394 | 0.267 | 0.073 | 0.128 | 0.185 | |
| P-DEBIASYear=20242026.06 | 0.199 | 0.458 | 0.367 | 0.05 | 0.091 | 0.109 | |
| BESTPERSONAYear=20242026.06 | 0.184 | 0.496 | 0.4 | 0.077 | 0.096 | 0.156 | |
| P-DEFENSEYear=20252026.06 | 0.181 | 0.432 | 0.367 | 0.05 | 0.066 | 0.117 | |
| Qwen3-4B-Instruct-25072026.06 | 0.17 | 0.477 | 0.409 | 0.052 | 0.068 | 0.109 | |
| 2STEPPROMPTYear=20252026.06 | 0.155 | 0.494 | 0.423 | 0.074 | 0.071 | 0.149 |