List your company

Personalized Response Generation on AIME Explicit Preference 2025 (test)

69.9PrefTriAlign
15.92429.93743.9557.963Jun 1, 2026

Evaluation results

MethodLinks
TriAlign2026.0669.946.743.32.73.35.8
Qwen3-4B-Instruct-25072026.0669.132.826.73.66.110.9
SWIFTYear=20262026.0641.840307.21018
SFTYear=20222026.0637.536.1305.66.115.5
BESTPERSONAYear=20242026.0637.43326.75.36.316
P-DEFENSEYear=20252026.0633.93026.73.83.312.8
P-DEBIASYear=20242026.0633.130.926.72.54.28.2
2STEPPROMPTYear=20252026.061848.242.35.15.910.6