Divergence based Ranked Preference Alignment for LLMs

arXiv:2610.25588v1 Announce Type: new Abstract: Alignment of large language models typically relies on pairwise preference feedback, though ranked feedback can capture more comprehensive information across multiple responses within a single annotation. However, ranked-preference alignment is underexplored in practice compared to simpler pairwise approaches, potentially resulting in suboptimal behavior. Concurrently, recent research has conceptualized large language model alignment objectives as divergence estimators between aligned (preferred) and unaligned (less-preferred) response distributions, resulting in a family of objectives with an interpretable theoretical foundation and strong empirical performance. This work unifies these perspectives by introducing $f$-RPO, a $f$-divergence objective for ranked preference optimization that extends divergence-based alignment to ranked feedback and recovers pairwise preference alignment as a special case. Our theoretical analysis establishes rank-order consistency, showing that $f$-RPO preserves the underlying preference ordering at convergence. Empirical results indicate that total-variation and reverse-KL variants of the proposed objective consistently outperform ranked-alignment baselines in both ranking and generation quality across Qwen, Llama, & Gemma models ranging from 0.5B to 8B parameters.

Abstract:Alignment of large language models typically relies on pairwise preference feedback, though ranked feedback can capture more comprehensive information across multiple responses within a single annotation. However, ranked-preference alignment is underexplored in practice compared to simpler pairwise approaches, potentially resulting in suboptimal behavior. Concurrently, recent research has conceptualized large language model alignment objectives as divergence estimators between aligned (preferred) and unaligned (less-preferred) response distributions, resulting in a family of objectives with an interpretable theoretical foundation and strong empirical performance. This work unifies these perspectives by introducing $f$-RPO, a $f$-divergence objective for ranked preference optimization that extends divergence-based alignment to ranked feedback and recovers pairwise preference alignment as a special case. Our theoretical analysis establishes rank-order consistency, showing that $f$-RPO preserves the underlying preference ordering at convergence. Empirical results indicate that total-variation and reverse-KL variants of the proposed objective consistently outperform ranked-alignment baselines in both ranking and generation quality across Qwen, Llama, & Gemma models ranging from 0.5B to 8B parameters.

Machine Learning (cs.LG); Machine Learning (stat.ML)

arXiv:2610.25588 [cs.LG]

(or arXiv:2610.25588v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2610.25588 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

aiqbench analysis

Divergence based Ranked Preference Alignment for LLMs

Related coverage