# ADPO: Anchored Direct Preference Optimization

arXiv: https://arxiv.org/abs/2510.18913 · HTML: https://mlsanity.com/paper/2510.18913/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in LLM preference optimization methods, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 20 of its 34 references that are in the map.
- First posted on arXiv 21 Oct 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [LLM preference optimization methods](https://mlsanity.com/alignment/llm-preference-optimization-methods/)

## Builds on (most cited first, 20 of 20)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor](https://mlsanity.com/paper/1801.01290/) (2018)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Trust Region Policy Optimization](https://mlsanity.com/paper/1502.05477/) (2015)
- [HybridFlow: A Flexible and Efficient RLHF Framework](https://mlsanity.com/paper/2409.19256/) (2024)
- [A General Theoretical Paradigm to Understand Learning from Human Preferences](https://mlsanity.com/paper/2310.12036/) (2023)
- [Advantage-Weighted Regression: Simple and Scalable Off-Policy Reinforcement Learning](https://mlsanity.com/paper/1910.00177/) (2019)
- [AWAC: Accelerating Online Reinforcement Learning with Offline Datasets](https://mlsanity.com/paper/2006.09359/) (2020)
- [Reinforcement Learning with Deep Energy-Based Policies](https://mlsanity.com/paper/1702.08165/) (2017)
- [Iterative Preference Learning from Human Feedback: Bridging Theory and Practice for RLHF under KL-Constraint](https://mlsanity.com/paper/2312.11456/) (2023)
- [Maximum a Posteriori Policy Optimisation](https://mlsanity.com/paper/1806.06920/) (2018)
- [Direct Nash Optimization: Teaching Language Models to Self-Improve with General Preferences](https://mlsanity.com/paper/2404.03715/) (2024)
- [Calibrating Sequence likelihood Improves Conditional Language Generation](https://mlsanity.com/paper/2210.00045/) (2022)
- [Quantifying Differences in Reward Functions](https://mlsanity.com/paper/2006.13900/) (2020)
- [Preference Optimization for Combinatorial Optimization Problems](https://mlsanity.com/paper/2505.08735/) (2025)

## Built on by (most cited first, 3 of 3)

- [Orthogonalized Policy Optimization:Policy Optimization as Orthogonal Projection in Hilbert Space](https://mlsanity.com/paper/2601.12415/) (2026)
- [Group Orthogonalized Policy Optimization:Group Policy Optimization as Orthogonal Projection in Hilbert Space](https://mlsanity.com/paper/2602.21269/) (2026)
- [BA-DPO: Bias-Adjusted Direct Preference Optimization for Language Model Alignment](https://mlsanity.com/paper/2609.35044/) (2026)

## Abstract (arXiv)

We present Anchored Direct Preference Optimization (ADPO), a policy alignment method derived from first principles of KL-regularized reinforcement learning. Unlike standard approaches that treat the reference policy merely as a regularizer, we show that the optimal policy in reinforcement learning from human feedback inherently operates in a differential coordinate system, optimizing relative advantage in the form of log ratios rather than absolute probabilities. ADPO explicitly parameterizes this optimal structure through anchored logits, effectively decoupling response quality from prior popularity and creating an implicit trust region through curvature scaling. We show that this formulation unifies supervised fine-tuning, reinforcement learning, and ranking-based objectives under a single geometric perspective. Theoretically, ADPO resolves the probability smearing problem of supervised fine-tuning while avoiding the mode-seeking instability characteristic of reverse-KL methods. Empirically, the listwise ranking variant of ADPO achieves state-of-the-art performance on reasoning tasks, outperforming GRPO by 30.9 percent on Qwen3-1.7B and demonstrating superior robustness under distribution shift.