Relative Density Ratio Optimization for Stable and Statistically Consistent Model Alignment
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910105629360128 |
|---|---|
| author | Takahashi, Hiroshi Iwata, Tomoharu Kumagai, Atsutoshi Kanai, Sekitoshi Yamada, Masanori Nishida, Kosuke Shinoda, Kazutoshi |
| author_facet | Takahashi, Hiroshi Iwata, Tomoharu Kumagai, Atsutoshi Kanai, Sekitoshi Yamada, Masanori Nishida, Kosuke Shinoda, Kazutoshi |
| contents | Aligning language models with human preferences is essential for ensuring their safety and reliability. Although most existing approaches assume specific human preference models such as the Bradley-Terry model, this assumption may fail to accurately capture true human preferences, and consequently, these methods lack statistical consistency, i.e., the guarantee that language models converge to the true human preference as the number of samples increases. In contrast, direct density ratio optimization (DDRO) achieves statistical consistency without assuming any human preference models. DDRO models the density ratio between preferred and non-preferred data distributions using the language model, and then optimizes it via density ratio estimation. However, this density ratio is unstable and often diverges, leading to training instability of DDRO. In this paper, we propose a novel alignment method that is both stable and statistically consistent. Our approach is based on the relative density ratio between the preferred data distribution and a mixture of the preferred and non-preferred data distributions. Our approach is stable since this relative density ratio is bounded above and does not diverge. Moreover, it is statistically consistent and yields significantly tighter convergence guarantees than DDRO. We experimentally show its effectiveness with Qwen 2.5 and Llama 3. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_04410 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Relative Density Ratio Optimization for Stable and Statistically Consistent Model Alignment Takahashi, Hiroshi Iwata, Tomoharu Kumagai, Atsutoshi Kanai, Sekitoshi Yamada, Masanori Nishida, Kosuke Shinoda, Kazutoshi Machine Learning Artificial Intelligence Computation and Language Aligning language models with human preferences is essential for ensuring their safety and reliability. Although most existing approaches assume specific human preference models such as the Bradley-Terry model, this assumption may fail to accurately capture true human preferences, and consequently, these methods lack statistical consistency, i.e., the guarantee that language models converge to the true human preference as the number of samples increases. In contrast, direct density ratio optimization (DDRO) achieves statistical consistency without assuming any human preference models. DDRO models the density ratio between preferred and non-preferred data distributions using the language model, and then optimizes it via density ratio estimation. However, this density ratio is unstable and often diverges, leading to training instability of DDRO. In this paper, we propose a novel alignment method that is both stable and statistically consistent. Our approach is based on the relative density ratio between the preferred data distribution and a mixture of the preferred and non-preferred data distributions. Our approach is stable since this relative density ratio is bounded above and does not diverge. Moreover, it is statistically consistent and yields significantly tighter convergence guarantees than DDRO. We experimentally show its effectiveness with Qwen 2.5 and Llama 3. |
| title | Relative Density Ratio Optimization for Stable and Statistically Consistent Model Alignment |
| topic | Machine Learning Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2604.04410 |