RainbowPO: A Unified Framework for Combining Improvements in Preference Optimization

• URL: http://arxiv.org/abs/2410.04203v1
• Date: Sat, 5 Oct 2024 15:44:46 GMT
• Title: RainbowPO: A Unified Framework for Combining Improvements in Preference Optimization
• Authors: Hanyang Zhao, Genta Indra Winata, Anirban Das, Shi-Xiong Zhang, David D. Yao, Wenpin Tang, Sambit Sahu,
• Abstract summary: RainbowPO is a unified framework that categorizing key components into seven broad directions. We demonstrate that RainbowPO outperforms existing DPO variants. We provide insights to guide researchers in developing new DPO methods and assist practitioners in their implementations.
• Score: 22.45649373554474
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recently, numerous preference optimization algorithms have been introduced as extensions to the Direct Preference Optimization (DPO) family. While these methods have successfully aligned models with human preferences, there is a lack of understanding regarding the contributions of their additional components. Moreover, fair and consistent comparisons are scarce, making it difficult to discern which components genuinely enhance downstream performance. In this work, we propose RainbowPO, a unified framework that demystifies the effectiveness of existing DPO methods by categorizing their key components into seven broad directions. We integrate these components into a single cohesive objective, enhancing the performance of each individual element. Through extensive experiments, we demonstrate that RainbowPO outperforms existing DPO variants. Additionally, we provide insights to guide researchers in developing new DPO methods and assist practitioners in their implementations.

Related papers

• Reward-aware Preference Optimization: A Unified Mathematical Framework for Model Alignment [45.45508377432791]
  This paper introduces Reward-Aware Preference Optimization (RPO), a mathematical framework that unifies popular preference optimization techniques. RPO provides a structured approach to disentangle and systematically study the impact of various design choices. We propose a new experimental setup that enables the clean and direct ablation of such design choices.
  arXiv  Detail & Related papers  (2025-01-31T22:39:04Z)
• AlphaPO - Reward shape matters for LLM alignment [8.688476316386176]
  We introduce textbfAlphaPO, a new DAA that helps change the shape of the reward function beyond the standard log reward. Compared to SimPO, one of the best performing DAAs, AlphaPO leads to about 7% to 10% relative improvement in alignment performance.
  arXiv  Detail & Related papers  (2025-01-07T15:46:42Z)
• Many of Your DPOs are Secretly One: Attempting Unification Through Mutual Information [5.655057078073446]
  Post-alignment of large language models (LLMs) is critical in improving their utility, safety, and alignment with human intentions. Direct preference optimisation (DPO) has become one of the most widely used algorithms for achieving this alignment. This paper introduces a unifying framework inspired by mutual information, which proposes a new loss function with flexible priors.
  arXiv  Detail & Related papers  (2025-01-02T21:31:38Z)
• $f$-PO: Generalizing Preference Optimization with $f$-divergence Minimization [54.94545757220999]
  $f$-PO is a novel framework that generalizes and extends existing approaches. We conduct experiments on state-of-the-art language models using benchmark datasets.
  arXiv  Detail & Related papers  (2024-10-29T02:11:45Z)
• Accelerated Preference Optimization for Large Language Model Alignment [60.22606527763201]
  Reinforcement Learning from Human Feedback (RLHF) has emerged as a pivotal tool for aligning large language models (LLMs) with human preferences. Direct Preference Optimization (DPO) formulates RLHF as a policy optimization problem without explicitly estimating the reward function. We propose a general Accelerated Preference Optimization (APO) framework, which unifies many existing preference optimization algorithms.
  arXiv  Detail & Related papers  (2024-10-08T18:51:01Z)
• End-to-End Learnable Item Tokenization for Generative Recommendation [51.82768744368208]
  We propose ETEGRec, a novel End-To-End Generative Recommender by seamlessly integrating item tokenization and generative recommendation. Our framework is developed based on the dual encoder-decoder architecture, which consists of an item tokenizer and a generative recommender.
  arXiv  Detail & Related papers  (2024-09-09T12:11:53Z)
• 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)
• Learning k-Determinantal Point Processes for Personalized Ranking [13.677246792673564]
  We present a new optimization criterion LkP based on set probability comparison for personalized ranking. LkP is broadly applicable, and when applied to existing recommendation models it also yields strong performance improvements.
  arXiv  Detail & Related papers  (2024-06-23T02:24:50Z)
• D2PO: Discriminator-Guided DPO with Response Evaluation Models [63.71853401569461]
  We propose D2PO, discriminator-guided DPO, for the online setting where preferences are being collected throughout learning. As we collect gold preferences, we use these not only to train our policy, but to train a discriminative response evaluation model to silver-label even more synthetic data for policy training. We show conditions under which silver labeling is most helpful: it is most effective when training the policy with DPO, outperforming traditional PPO, and benefits from maintaining a separate discriminator from the policy model.
  arXiv  Detail & Related papers  (2024-05-02T17:44:41Z)
• Relative Preference Optimization: Enhancing LLM Alignment through Contrasting Responses across Identical and Diverse Prompts [95.09994361995389]
  Relative Preference Optimization (RPO) is designed to discern between more and less preferred responses derived from both identical and related prompts. RPO has demonstrated a superior ability to align large language models with user preferences and to improve their adaptability during the training process.
  arXiv  Detail & Related papers  (2024-02-12T22:47:57Z)

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.