Alignment as Distribution Learning: Your Preference Model is Explicitly a Language Model

• URL: http://arxiv.org/abs/2506.01523v1
• Date: Mon, 02 Jun 2025 10:36:31 GMT
• Title: Alignment as Distribution Learning: Your Preference Model is Explicitly a Language Model
• Authors: Jihun Yun, Juno Kim, Jongho Park, Junhyuck Kim, Jongha Jon Ryu, Jaewoong Cho, Kwang-Sung Jun,
• Abstract summary: We argue that alignment via reinforcement learning from human feedback lacks theoretical justification and incentivizes deterministic solutions.<n>We propose three principled learning objectives: preference maximum likelihood estimation, preference distillation, and reverse KL minimization.<n>We empirically demonstrate that our distribution learning framework, especially preference distillation, consistently outperforms or matches the performances of RLHF and DPO.
• Score: 12.063078727764045
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Alignment via reinforcement learning from human feedback (RLHF) has become the dominant paradigm for controlling the quality of outputs from large language models (LLMs). However, when viewed as `loss + regularization,' the standard RLHF objective lacks theoretical justification and incentivizes degenerate, deterministic solutions, an issue that variants such as Direct Policy Optimization (DPO) also inherit. In this paper, we rethink alignment by framing it as \emph{distribution learning} from pairwise preference feedback by explicitly modeling how information about the target language model bleeds through the preference data. This explicit modeling leads us to propose three principled learning objectives: preference maximum likelihood estimation, preference distillation, and reverse KL minimization. We theoretically show that all three approaches enjoy strong non-asymptotic $O(1/n)$ convergence to the target language model, naturally avoiding degeneracy and reward overfitting. Finally, we empirically demonstrate that our distribution learning framework, especially preference distillation, consistently outperforms or matches the performances of RLHF and DPO across various tasks and models.

Related papers

• Understanding the Performance Gap in Preference Learning: A Dichotomy of RLHF and DPO [51.22869332661607]
  We decompose the performance gap between reinforcement learning from human feedback and direct preference optimization under a representation gap.<n>We show that RLHF, DPO, or online DPO can outperform one another depending on the type of model mis-specifications.
  arXiv  Detail & Related papers  (2025-05-26T09:54:02Z)
• Self-Boost via Optimal Retraining: An Analysis via Approximate Message Passing [58.52119063742121]
  Retraining a model using its own predictions together with the original, potentially noisy labels is a well-known strategy for improving the model performance.<n>This paper addresses the question of how to optimally combine the model's predictions and the provided labels.<n>Our main contribution is the derivation of the Bayes optimal aggregator function to combine the current model's predictions and the given labels.
  arXiv  Detail & Related papers  (2025-05-21T07:16:44Z)
• Self-supervised Preference Optimization: Enhance Your Language Model with Preference Degree Awareness [27.43137305486112]
  We propose a novel Self-supervised Preference Optimization (SPO) framework, which constructs a self-supervised preference degree loss combined with the alignment loss. The results demonstrate that SPO can be seamlessly integrated with existing preference optimization methods to achieve state-of-the-art performance.
  arXiv  Detail & Related papers  (2024-09-26T12:37:26Z)
• Would I Lie To You? Inference Time Alignment of Language Models using Direct Preference Heads [11.254305578659002]
  We introduce Direct Preference Heads (DPH), a fine-tuning framework that enables LMs to learn human preference signals through an auxiliary reward head without directly affecting the output distribution of the language modeling head. We evaluate our models on GLUE, RACE, and the GPT4All evaluation suite and demonstrate that our method produces models which achieve higher scores than those fine-tuned with Supervised Fine-Tuning (SFT) or Direct Preference Optimization (DPO) alone.
  arXiv  Detail & Related papers  (2024-05-30T13:38:52Z)
• Preference Learning Algorithms Do Not Learn Preference Rankings [62.335733662381884]
  We study the conventional wisdom that preference learning trains models to assign higher likelihoods to more preferred outputs than less preferred outputs. We find that most state-of-the-art preference-tuned models achieve a ranking accuracy of less than 60% on common preference datasets.
  arXiv  Detail & Related papers  (2024-05-29T21:29:44Z)
• Robust Preference Optimization through Reward Model Distillation [68.65844394615702]
  Direct Preference Optimization (DPO) is a popular offline alignment method that trains a policy directly on preference data.<n>We analyze this phenomenon and use distillation to get a better proxy for the true preference distribution over generation pairs.<n>Our results show that distilling from such a family of reward models leads to improved robustness to distribution shift in preference annotations.
  arXiv  Detail & Related papers  (2024-05-29T17:39:48Z)
• RLHF from Heterogeneous Feedback via Personalization and Preference Aggregation [24.374185140811115]
  Reinforcement learning from human feedback (RLHF) has been an effective technique for aligning AI systems with human values. In this paper, we focus on addressing the issues due to the inherent heterogeneity in human preferences, as well as their potential strategic behavior in providing feedback. We propose two frameworks to address heterogeneous human feedback in principled ways: personalization-based one and aggregation-based one.
  arXiv  Detail & Related papers  (2024-04-30T23:57:23Z)
• MaxMin-RLHF: Alignment with Diverse Human Preferences [101.57443597426374]
  Reinforcement Learning from Human Feedback (RLHF) aligns language models to human preferences by employing a singular reward model derived from preference data.<n>We learn a mixture of preference distributions via an expectation-maximization algorithm to better represent diverse human preferences.<n>Our algorithm achieves an average improvement of more than 16% in win-rates over conventional RLHF algorithms.
  arXiv  Detail & Related papers  (2024-02-14T03:56:27Z)
• Secrets of RLHF in Large Language Models Part II: Reward Modeling [134.97964938009588]
  We introduce a series of novel methods to mitigate the influence of incorrect and ambiguous preferences in the dataset. We also introduce contrastive learning to enhance the ability of reward models to distinguish between chosen and rejected responses.
  arXiv  Detail & Related papers  (2024-01-11T17:56:59Z)
• Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization [76.09576643028362]
  We present Multi-Objective Direct Preference Optimization (MODPO) for multiple alignment objectives. MODPO folds language modeling directly into reward modeling, training language models as implicit collective reward models. It theoretically yields the same optimal solutions as MORLHF but is practically more stable and efficient.
  arXiv  Detail & Related papers  (2023-10-05T17:35:26Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.