Anyprefer: An Agentic Framework for Preference Data Synthesis

• URL: http://arxiv.org/abs/2504.19276v1
• Date: Sun, 27 Apr 2025 15:21:59 GMT
• Title: Anyprefer: An Agentic Framework for Preference Data Synthesis
• Authors: Yiyang Zhou, Zhaoyang Wang, Tianle Wang, Shangyu Xing, Peng Xia, Bo Li, Kaiyuan Zheng, Zijian Zhang, Zhaorun Chen, Wenhao Zheng, Xuchao Zhang, Chetan Bansal, Weitong Zhang, Ying Wei, Mohit Bansal, Huaxiu Yao,
• Abstract summary: We propose Anyprefer, a framework designed to synthesize high-quality preference data for aligning the target model.<n> external tools are introduced to assist the judge model in accurately rewarding the target model's responses.<n>The synthesized data is compiled into a new preference dataset, Anyprefer-V1, consisting of 58K high-quality preference pairs.
• Score: 62.3856754548222
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: High-quality preference data is essential for aligning foundation models with human values through preference learning. However, manual annotation of such data is often time-consuming and costly. Recent methods often adopt a self-rewarding approach, where the target model generates and annotates its own preference data, but this can lead to inaccuracies since the reward model shares weights with the target model, thereby amplifying inherent biases. To address these issues, we propose Anyprefer, a framework designed to synthesize high-quality preference data for aligning the target model. Anyprefer frames the data synthesis process as a cooperative two-player Markov Game, where the target model and the judge model collaborate together. Here, a series of external tools are introduced to assist the judge model in accurately rewarding the target model's responses, mitigating biases in the rewarding process. In addition, a feedback mechanism is introduced to optimize prompts for both models, enhancing collaboration and improving data quality. The synthesized data is compiled into a new preference dataset, Anyprefer-V1, consisting of 58K high-quality preference pairs. Extensive experiments show that Anyprefer significantly improves model alignment performance across four main applications, covering 21 datasets, achieving average improvements of 18.55% in five natural language generation datasets, 3.66% in nine vision-language understanding datasets, 30.05% in three medical image analysis datasets, and 16.00% in four visuo-motor control tasks.

Related papers

• Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models [7.61977883644433]
  We propose PRRC to evaluate data quality across Professionalism, Readability, Reasoning, and Cleanliness. We introduce Meta-rater, a multi-dimensional data selection method that integrates these dimensions with existing quality metrics through learned optimal weightings. Experiments demonstrate that Meta-rater doubles convergence speed for 1.3B parameter models and improves downstream task performance by 3.23, with scalable benefits observed in 3.3B models trained on 100B tokens.
  arXiv  Detail & Related papers  (2025-04-19T06:12:33Z)
• More is Less: The Pitfalls of Multi-Model Synthetic Preference Data in DPO Safety Alignment [80.04449725137177]
  Direct Preference Optimization (DPO) has emerged as a simple, yet effective alternative to reinforcement learning from human feedback. Our study reveals a striking, safety-specific phenomenon associated with DPO alignment. Using solely self-generated responses for both chosen and rejected pairs significantly outperforms configurations that incorporate responses from stronger models.
  arXiv  Detail & Related papers  (2025-04-03T00:36:40Z)
• Evaluating Sample Utility for Data Selection by Mimicking Model Weights [12.056542160711718]
  Foundation models are trained on large-scale web-crawled datasets, which often contain noise, biases, and irrelevant information.<n>We propose an efficient, model-based approach using the Mimic Score, a new data quality metric.<n>We develop Grad-Mimic, a framework that prioritizes samples for learning, creates effective filters, and automates data selection.
  arXiv  Detail & Related papers  (2025-01-12T04:28:14Z)
• Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedback [87.37721254914476]
  We introduce a routing framework that combines inputs from humans and LMs to achieve better annotation quality. We train a performance prediction model to predict a reward model's performance on an arbitrary combination of human and LM annotations. We show that the selected hybrid mixture achieves better reward model performance compared to using either one exclusively.
  arXiv  Detail & Related papers  (2024-10-24T20:04:15Z)
• Automated Filtering of Human Feedback Data for Aligning Text-to-Image Diffusion Models [36.84880190385986]
  Fine-tuning text-to-image diffusion models with human feedback often suffers from slow convergence due to the large size and noise present in feedback datasets.<n>We propose FiFA, a novel automated data filtering algorithm designed to enhance the fine-tuning of diffusion models using human feedback datasets.<n> Experimental results show that FiFA significantly enhances training stability and achieves better performance, being preferred by humans 17% more.
  arXiv  Detail & Related papers  (2024-10-14T05:18:07Z)
• Beyond Bradley-Terry Models: A General Preference Model for Language Model Alignment [51.14207112118503]
  We introduce preference embedding, an approach that embeds responses into a latent space to capture preferences efficiently.<n>We also propose preference score-based General Preference Optimization (GPO), which generalizes reward-based reinforcement learning from human feedback.<n>Our method may enhance the alignment of foundation models with nuanced human values.
  arXiv  Detail & Related papers  (2024-10-03T04:22:55Z)
• Towards Data-Centric RLHF: Simple Metrics for Preference Dataset Comparison [9.324894567200582]
  We systematically study preference datasets through three perspectives: scale, label noise, and information content. Our work is a first step towards a data-centric approach to alignment by providing perspectives that aid in training efficiency and iterative data collection for RLHF.
  arXiv  Detail & Related papers  (2024-09-15T03:55:03Z)
• Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment [57.03947082589616]
  Large Language Models (LLMs) are often aligned using contrastive alignment objectives and preference pair datasets. We study this and find that preference data gives a better learning signal when the underlying responses are contrastive. We introduce Contrastive Learning from AI Revisions (CLAIR), a data-creation method which leads to more contrastive preference pairs. Our best model, trained on 32K CLAIR preferences with APO, improves Llama-3-8B-Instruct by 7.65%, closing the gap with GPT4-turbo by 45%.
  arXiv  Detail & Related papers  (2024-08-12T16:24:51Z)
• Dataset Regeneration for Sequential Recommendation [69.93516846106701]
  We propose a data-centric paradigm for developing an ideal training dataset using a model-agnostic dataset regeneration framework called DR4SR. To demonstrate the effectiveness of the data-centric paradigm, we integrate our framework with various model-centric methods and observe significant performance improvements across four widely adopted datasets.
  arXiv  Detail & Related papers  (2024-05-28T03:45:34Z)
• 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)

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.