Relative density ratio optimization stabilizes direct density ratio estimation for language model alignment while preserving statistical consistency without assuming a Bradley-Terry preference model.
Energy-based preference model offers better offline alignment than the bradley-terry preference model.arXiv preprint arXiv:2412.13862
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Relative Density Ratio Optimization for Stable and Statistically Consistent Model Alignment
Relative density ratio optimization stabilizes direct density ratio estimation for language model alignment while preserving statistical consistency without assuming a Bradley-Terry preference model.