A density estimation perspective on learning from pairwise human preferences

• URL: http://arxiv.org/abs/2311.14115v3
• Date: Wed, 10 Jan 2024 16:11:32 GMT
• Title: A density estimation perspective on learning from pairwise human preferences
• Authors: Vincent Dumoulin, Daniel D. Johnson, Pablo Samuel Castro, Hugo Larochelle, Yann Dauphin
• Abstract summary: We show that for a family of generative processes defined via preference behavior distribution equations, training a reward function on pairwise preferences effectively models an annotator's implicit preference distribution. We discuss and present findings on "annotator misspecification" -- failure cases where wrong modeling assumptions are made about annotator behavior, resulting in poorly-adapted models.
• Score: 32.64330423345252
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Learning from human feedback (LHF) -- and in particular learning from pairwise preferences -- has recently become a crucial ingredient in training large language models (LLMs), and has been the subject of much research. Most recent works frame it as a reinforcement learning problem, where a reward function is learned from pairwise preference data and the LLM is treated as a policy which is adapted to maximize the rewards, often under additional regularization constraints. We propose an alternative interpretation which centers on the generative process for pairwise preferences and treats LHF as a density estimation problem. We provide theoretical and empirical results showing that for a family of generative processes defined via preference behavior distribution equations, training a reward function on pairwise preferences effectively models an annotator's implicit preference distribution. Finally, we discuss and present findings on "annotator misspecification" -- failure cases where wrong modeling assumptions are made about annotator behavior, resulting in poorly-adapted models -- suggesting that approaches that learn from pairwise human preferences could have trouble learning from a population of annotators with diverse viewpoints.

Related papers

• Calibrated Multi-Preference Optimization for Aligning Diffusion Models [92.90660301195396]
  Calibrated Preference Optimization (CaPO) is a novel method to align text-to-image (T2I) diffusion models. CaPO incorporates the general preference from multiple reward models without human annotated data. Experimental results show that CaPO consistently outperforms prior methods.
  arXiv  Detail & Related papers  (2025-02-04T18:59:23Z)
• Disentangling Length Bias In Preference Learning Via Response-Conditioned Modeling [87.17041933863041]
  We introduce a Response-conditioned Bradley-Terry (Rc-BT) model that enhances the reward model's capability in length bias mitigating and length instruction following. We also propose the Rc-DPO algorithm to leverage the Rc-BT model for direct policy optimization (DPO) of large language models.
  arXiv  Detail & Related papers  (2025-02-02T14:50:25Z)
• 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)
• Personalizing Reinforcement Learning from Human Feedback with Variational Preference Learning [12.742158403867002]
  Reinforcement Learning from Human Feedback is a powerful paradigm for aligning foundation models to human values and preferences. Current RLHF techniques cannot account for the naturally occurring differences in individual human preferences across a diverse population. We develop a class of multimodal RLHF methods to address the need for pluralistic alignment.
  arXiv  Detail & Related papers  (2024-08-19T15:18:30Z)
• Bridging and Modeling Correlations in Pairwise Data for Direct Preference Optimization [75.1240295759264]
  We propose an effective framework for Bridging and Modeling Correlations in pairwise data, named BMC. We increase the consistency and informativeness of the pairwise preference signals through targeted modifications. We identify that DPO alone is insufficient to model these correlations and capture nuanced variations.
  arXiv  Detail & Related papers  (2024-08-14T11:29:47Z)
• 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. We learn a mixture of preference distributions via an expectation-maximization algorithm to better represent diverse human preferences. 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)
• Models of human preference for learning reward functions [80.39289349661364]
  We learn the reward function from human-generated preferences between pairs of trajectory segments. We find this assumption to be flawed and propose modeling human preferences as informed by each segment's regret. Our proposed regret preference model better predicts real human preferences and also learns reward functions from these preferences that lead to policies that are better human-aligned.
  arXiv  Detail & Related papers  (2022-06-05T17:58:02Z)

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.