Accelerating Direct Preference Optimization with Prefix Sharing

• URL: http://arxiv.org/abs/2410.20305v2
• Date: Wed, 30 Oct 2024 18:02:43 GMT
• Title: Accelerating Direct Preference Optimization with Prefix Sharing
• Abstract summary: We introduce prefix sharing for preference tuning, a novel technique that processes chosen and rejected responses as one sequence with a shared prefix. Our method achieves $1.1$-$1.5times$ improvement in training throughput on popular DPO datasets, without any effect on convergence. Our work contributes to making preference-based fine-tuning more accessible for a wider range of applications and model sizes.
• Score: 0.9194662383948566
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Offline paired preference optimization algorithms have become a popular approach for fine-tuning on preference data, outperforming traditional supervised fine-tuning in various tasks. However, traditional implementations often involve redundant computations, especially for tasks with long shared prompts. We introduce prefix sharing for preference tuning, a novel technique that processes chosen and rejected responses as one sequence with a shared prefix. To prevent cross-response contamination, we use a custom block-sparse attention mask. Our method achieves $1.1$-$1.5\times$ improvement in training throughput on popular DPO datasets, without any effect on convergence. When combined with sequence packing, we observe consistent $1.3$-$1.6\times$ speedups, benefiting even datasets with smaller sequence lengths. While we focus on Direct Preference Optimization (DPO), our approach is applicable to other paired preference tuning methods. By enhancing computational efficiency, our work contributes to making preference-based fine-tuning more accessible for a wider range of applications and model sizes. We open-source our code at https://github.com/frankxwang/dpo-prefix-sharing.

Related papers

• Intrinsic Mutual Information as a Modulator for Preference Optimization [21.430166597980804]
  We propose RMiPO, a lightweight and efficient framework for offline preference optimization.<n>We show that RMiPO achieves consistently superior performance over existing methods while reducing training overhead by more than 15%.
  arXiv  Detail & Related papers  (2026-04-27T05:22:54Z)
• GroupDPO: Memory efficient Group-wise Direct Preference Optimization [29.469340103121144]
  We introduce a memory-efficient group-wise preference optimization algorithm that preserves gradients while decoupling samples during backpropagation.<n>Across both offline and online alignment settings, we show that leveraging multiple responses consistently outperforms single-pair training.
  arXiv  Detail & Related papers  (2026-04-17T00:56:59Z)
• Direct Preference Optimization with Unobserved Preference Heterogeneity: The Necessity of Ternary Preferences [14.686788596611246]
  Reinforcement Learning from Human Feedback (RLHF) has become central to aligning large language models with human values.<n>Recent alternatives such as Direct Preference Optimization (DPO) simplify this pipeline by directly optimizing on preferences.<n>We propose a theoretical and algorithmic framework for fairness and personalization for diverse users in generative model alignment.
  arXiv  Detail & Related papers  (2025-10-17T15:00:40Z)
• Adaptive Preference Optimization with Uncertainty-aware Utility Anchor [33.74005997646761]
  offline preference optimization methods are efficient for large language models (LLMs) alignment.<n>We propose a general framework for offline preference optimization methods, which introduces an anchoring function to estimate the uncertainties brought from preference data annotation.<n>Our method enables training even in scenarios where the data is unpaired, significantly enhancing data utilization efficiency.
  arXiv  Detail & Related papers  (2025-09-03T10:20:08Z)
• Listwise Preference Alignment Optimization for Tail Item Recommendation [29.20780001894192]
  We propose LPO4Rec, which extends the Bradley-Terry model from pairwise comparison to listwise comparison.<n>Specifically, we derive a closed form optimal policy to enable more efficient and effective training without explicit reward modeling.<n>Our experiments on three public datasets show that our method outperforms 10 baselines by a large margin.
  arXiv  Detail & Related papers  (2025-07-03T03:08:23Z)
• Smoothed Preference Optimization via ReNoise Inversion for Aligning Diffusion Models with Varied Human Preferences [13.588231827053923]
  Direct Preference Optimization (DPO) aligns text-to-image (T2I) generation models with human preferences using pairwise preference data.<n>We propose SmPO-Diffusion, a novel method for modeling preference distributions to improve the DPO objective.<n>Our approach effectively mitigates issues of excessive optimization and objective misalignment present in existing methods.
  arXiv  Detail & Related papers  (2025-06-03T09:47:22Z)
• Earlier Tokens Contribute More: Learning Direct Preference Optimization From Temporal Decay Perspective [22.248134630764497]
  We propose an enhanced preference optimization method that incorporates a temporal decay factor controlled by a gamma parameter. Our approach mitigates overfitting to less pertinent data and remains responsive to evolving human preferences.
  arXiv  Detail & Related papers  (2025-02-20T07:53:11Z)
• Smoothed Normalization for Efficient Distributed Private Optimization [54.197255548244705]
  Federated learning enables machine learning models with privacy of participants. There is no differentially private distributed method for training, non-feedback problems. We introduce a new distributed algorithm $alpha$-$sf NormEC$ with provable convergence guarantees.
  arXiv  Detail & Related papers  (2025-02-19T07:10:32Z)
• 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)
• Multi-Preference Optimization: Generalizing DPO via Set-Level Contrasts [17.243429150450886]
  We propose $textbfMulti-Preference Optimization (MPO) to optimize over entire sets of responses.<n>MPO employs deviation-based weighting, which emphasizes outlier responses that deviate most from the mean reward.<n>We theoretically prove that MPO reduces alignment bias at a rate of $mathcalOleft(frac1sqrtnright)$ with respect to the number of responses per query.
  arXiv  Detail & Related papers  (2024-12-05T21:50:22Z)
• Towards Improved Preference Optimization Pipeline: from Data Generation to Budget-Controlled Regularization [14.50339880957898]
  We aim to improve the preference optimization pipeline by taking a closer look at preference data generation and training regularization techniques. For preference data generation, we propose an iterative pairwise ranking mechanism that derives preference ranking of completions using pairwise comparison signals. For training regularization, we observe that preference optimization tends to achieve better convergence when the LLM predicted likelihood of preferred samples gets slightly reduced.
  arXiv  Detail & Related papers  (2024-11-07T23:03:11Z)
• Adaptive Preference Scaling for Reinforcement Learning with Human Feedback [103.36048042664768]
  Reinforcement learning from human feedback (RLHF) is a prevalent approach to align AI systems with human values. We propose a novel adaptive preference loss, underpinned by distributionally robust optimization (DRO) Our method is versatile and can be readily adapted to various preference optimization frameworks.
  arXiv  Detail & Related papers  (2024-06-04T20:33:22Z)
• Direct Preference Optimization With Unobserved Preference Heterogeneity: The Necessity of Ternary Preferences [14.686788596611246]
  Reinforcement Learning from Human Feedback (RLHF) has become central to aligning large language models with human values.<n>Recent alternatives such as Direct Preference Optimization (DPO) simplify this pipeline by directly optimizing on preferences.<n>We propose a theoretical and algorithmic framework for fairness and personalization for diverse users in generative model alignment.
  arXiv  Detail & Related papers  (2024-05-23T21:25:20Z)
• Comparing Bad Apples to Good Oranges: Aligning Large Language Models via Joint Preference Optimization [105.3612692153615]
  A common technique for aligning large language models (LLMs) relies on acquiring human preferences. We propose a new axis that is based on eliciting preferences jointly over the instruction-response pairs. We find that joint preferences over instruction and response pairs can significantly enhance the alignment of LLMs.
  arXiv  Detail & Related papers  (2024-03-31T02:05:40Z)
• 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)
• Voronoi Candidates for Bayesian Optimization [2.7309692684728617]
  Many practical BO methods, particularly in high dimension, eschew a formal, continuous optimization of the acquisition function. We propose to use candidates which lie on the boundary of the Voronoi tessellation of the current design points, so they are equidistant to two or more of them. We discuss strategies for efficient implementation by directly sampling the Voronoi boundary without explicitly generating the tessellation.
  arXiv  Detail & Related papers  (2024-02-07T14:47:13Z)
• Oracle-Efficient Pessimism: Offline Policy Optimization in Contextual Bandits [82.28442917447643]
  We present the first general oracle-efficient algorithm for pessimistic OPO. We obtain statistical guarantees analogous to those for prior pessimistic approaches. We show advantage over unregularized OPO across a wide range of configurations.
  arXiv  Detail & Related papers  (2023-06-13T17:29:50Z)
• Pre-training helps Bayesian optimization too [49.28382118032923]
  We seek an alternative practice for setting functional priors. In particular, we consider the scenario where we have data from similar functions that allow us to pre-train a tighter distribution a priori. Our results show that our method is able to locate good hyper parameters at least 3 times more efficiently than the best competing methods.
  arXiv  Detail & Related papers  (2022-07-07T04:42:54Z)
• You May Not Need Ratio Clipping in PPO [117.03368180633463]
  Proximal Policy Optimization (PPO) methods learn a policy by iteratively performing multiple mini-batch optimization epochs of a surrogate objective with one set of sampled data. Ratio clipping PPO is a popular variant that clips the probability ratios between the target policy and the policy used to collect samples. We show in this paper that such ratio clipping may not be a good option as it can fail to effectively bound the ratios. We show that ESPO can be easily scaled up to distributed training with many workers, delivering strong performance as well.
  arXiv  Detail & Related papers  (2022-01-31T20:26:56Z)
• Efficient Multi-Objective Optimization for Deep Learning [2.0305676256390934]
  Multi-objective optimization (MOO) is a prevalent challenge for Deep Learning. There exists no scalable MOO solution for truly deep neural networks.
  arXiv  Detail & Related papers  (2021-03-24T17:59:42Z)

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.