Mitigating Reward Over-optimization in Direct Alignment Algorithms with Importance Sampling

• URL: http://arxiv.org/abs/2506.08681v2
• Date: Wed, 11 Jun 2025 06:36:33 GMT
• Title: Mitigating Reward Over-optimization in Direct Alignment Algorithms with Importance Sampling
• Authors: Phuc Minh Nguyen, Ngoc-Hieu Nguyen, Duy H. M. Nguyen, Anji Liu, An Mai, Binh T. Nguyen, Daniel Sonntag, Khoa D. Doan,
• Abstract summary: Direct Alignment Algorithms (DAAs) have emerged as alternatives to the standard Reinforcement Learning from Human Feedback (RLHF)<n>These methods are more susceptible to over-optimization, in which the model drifts away from the reference policy, leading to degraded performance as training progresses.<n>This paper proposes a novel importance-sampling approach to mitigate the over-optimization problem of offline DAAs.
• Score: 13.917799959981185
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Direct Alignment Algorithms (DAAs) such as Direct Preference Optimization (DPO) have emerged as alternatives to the standard Reinforcement Learning from Human Feedback (RLHF) for aligning large language models (LLMs) with human values. However, these methods are more susceptible to over-optimization, in which the model drifts away from the reference policy, leading to degraded performance as training progresses. This paper proposes a novel importance-sampling approach to mitigate the over-optimization problem of offline DAAs. This approach, called (IS-DAAs), multiplies the DAA objective with an importance ratio that accounts for the reference policy distribution. IS-DAAs additionally avoid the high variance issue associated with importance sampling by clipping the importance ratio to a maximum value. Our extensive experiments demonstrate that IS-DAAs can effectively mitigate over-optimization, especially under low regularization strength, and achieve better performance than other methods designed to address this problem. Our implementations are provided publicly at this link.

Related papers

• Towards Bridging the Reward-Generation Gap in Direct Alignment Algorithms [6.844727437767206]
  Direct Alignment Algorithms (DAAs) have emerged as efficient alternatives to Reinforcement Learning from Human Feedback (RLHF) algorithms.<n>DAAs suffer from a misalignment between optimization objectives during training and actual generation performance during inference.<n>We introduce a simple yet effective approach called Prefix-Oriented Equal-length Training (POET)<n>POET truncates both preferred and dispreferred responses to match the shorter one's length.
  arXiv  Detail & Related papers  (2025-06-11T07:02:18Z)
• Token-Importance Guided Direct Preference Optimization [2.230951739798399]
  We propose a Token-Importance Guided Direct Preference Optimization (TI-DPO) to ensure that large language models generate outputs aligned with human preferences.<n> Experimental results show that TI-DPO achieves higher accuracy and stronger generative diversity, providing more stable and computationally efficient solutions.
  arXiv  Detail & Related papers  (2025-05-26T08:11:24Z)
• Leveraging Robust Optimization for LLM Alignment under Distribution Shifts [52.983390470606146]
  Preference alignment methods are increasingly critical for steering large language models to generate outputs consistent with human values.<n>We propose a novel distribution-aware optimization framework that improves preference alignment despite such shifts.
  arXiv  Detail & Related papers  (2025-04-08T09:14:38Z)
• A Simple and Effective Reinforcement Learning Method for Text-to-Image Diffusion Fine-tuning [61.403275660120606]
  Reinforcement learning (RL)-based fine-tuning has emerged as a powerful approach for aligning diffusion models with black-box objectives.<n>We propose leave-one-out PPO (LOOP), a novel RL for diffusion fine-tuning method.<n>Our results demonstrate that LOOP effectively improves diffusion models on various black-box objectives, and achieves a better balance between computational efficiency and performance.
  arXiv  Detail & Related papers  (2025-03-02T13:43:53Z)
• On-the-fly Preference Alignment via Principle-Guided Decoding [27.50204023448716]
  We introduce On-the-fly Preference Alignment via Principle-Guided Decoding (OPAD) to align model outputs with human preferences during inference.<n>OPAD achieves competitive or superior performance in both general and personalized alignment tasks.
  arXiv  Detail & Related papers  (2025-02-20T02:23:09Z)
• 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.<n>We increase the consistency and informativeness of the pairwise preference signals through targeted modifications.<n>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 Reward and Policy Jointly from Demonstration and Preference Improves Alignment [58.049113055986375]
  We develop a single stage approach named Alignment with Integrated Human Feedback (AIHF) to train reward models and the policy.<n>The proposed approach admits a suite of efficient algorithms, which can easily reduce to, and leverage, popular alignment algorithms.<n>We demonstrate the efficiency of the proposed solutions with extensive experiments involving alignment problems in LLMs and robotic control problems in MuJoCo.
  arXiv  Detail & Related papers  (2024-06-11T01:20:53Z)
• Provably Mitigating Overoptimization in RLHF: Your SFT Loss is Implicitly an Adversarial Regularizer [52.09480867526656]
  We identify the source of misalignment as a form of distributional shift and uncertainty in learning human preferences.<n>To mitigate overoptimization, we first propose a theoretical algorithm that chooses the best policy for an adversarially chosen reward model.<n>Using the equivalence between reward models and the corresponding optimal policy, the algorithm features a simple objective that combines a preference optimization loss and a supervised learning loss.
  arXiv  Detail & Related papers  (2024-05-26T05:38:50Z)
• Towards Efficient Exact Optimization of Language Model Alignment [93.39181634597877]
  Direct preference optimization (DPO) was proposed to directly optimize the policy from preference data. We show that DPO derived based on the optimal solution of problem leads to a compromised mean-seeking approximation of the optimal solution in practice. We propose efficient exact optimization (EXO) of the alignment objective.
  arXiv  Detail & Related papers  (2024-02-01T18:51:54Z)
• Adversarial Preference Optimization: Enhancing Your Alignment via RM-LLM Game [31.66896160733569]
  We propose an Adversarial Preference Optimization (APO) framework to target more efficient human preference optimization. We find the proposed adversarial training framework further enhances existing alignment baselines in terms of LLM helpfulness and harmlessness.
  arXiv  Detail & Related papers  (2023-11-14T10:10:31Z)

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.