Pluralistic Off-policy Evaluation and Alignment

• URL: http://arxiv.org/abs/2509.19333v1
• Date: Mon, 15 Sep 2025 01:57:49 GMT
• Title: Pluralistic Off-policy Evaluation and Alignment
• Authors: Chengkai Huang, Junda Wu, Zhouhang Xie, Yu Xia, Rui Wang, Tong Yu, Subrata Mitra, Julian McAuley, Lina Yao,
• Abstract summary: We propose POPE, the first framework for offline pluralistic preference evaluation and alignment in LLMs.<n>POPE includes a unified reward function that combines a collaborative utility component derived from human preference signals and a diversity component inspired by entropy-based coverage measures.<n> Empirical results demonstrate that POPE efficiently enhances pluralistic response generation and maintains the models' general capabilities on downstream tasks.
• Score: 47.35585359400588
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Personalized preference alignment for LLMs with diverse human preferences requires evaluation and alignment methods that capture pluralism. Most existing preference alignment datasets are logged under policies that differ substantially from the evaluated LLMs, and existing off-policy estimators focus solely on overall utility while ignoring preference pluralism. Extending Off-Policy Evaluation (OPE) to pluralistic preference alignment, therefore, remains an open question. Thus, we propose the Pluralistic Off-Policy Evaluation (POPE), the first framework for offline pluralistic preference evaluation and alignment in LLMs. POPE includes a unified reward function that combines (1) a collaborative utility component derived from human preference signals (e.g., upvotes or relevance scores) and (2) a diversity component inspired by entropy-based coverage measures, together reflecting pluralistic alignment. Furthermore, to estimate this reward from logged interactions, we derive decomposable inverse propensity scoring (IPS) estimators that separately evaluate relevance and diversity. Theoretically, we prove that our decomposed IPS estimators establish a lower bound on their variance. With the off-policy evaluated value function, we can directly enable off-policy optimization to further enhance pluralistic alignment. Empirical results demonstrate that POPE efficiently enhances pluralistic response generation and maintains the models' general capabilities on downstream tasks

Related papers

• Rethinking the Trust Region in LLM Reinforcement Learning [72.25890308541334]
  Proximal Policy Optimization (PPO) serves as the de facto standard algorithm for Large Language Models (LLMs)<n>We propose Divergence Proximal Policy Optimization (DPPO), which substitutes clipping with a more principled constraint.<n>DPPO achieves superior training and efficiency compared to existing methods, offering a more robust foundation for RL-based fine-tuning.
  arXiv  Detail & Related papers  (2026-02-04T18:59:04Z)
• GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization [133.27496265096445]
  We show how to apply Group Relative Policy Optimization under multi-reward setting without examining its suitability.<n>We then introduce Group reward-Decoupled Normalization Policy Optimization (GDPO), a new policy optimization method to resolve these issues.<n>GDPO consistently outperforms GRPO, demonstrating its effectiveness and generalizability for multi-reward reinforcement learning optimization.
  arXiv  Detail & Related papers  (2026-01-08T18:59:24Z)
• Importance Sampling for Multi-Negative Multimodal Direct Preference Optimization [68.64764778089229]
  We propose MISP-DPO, the first framework to incorporate multiple, semantically diverse negative images in multimodal DPO.<n>Our method embeds prompts and candidate images in CLIP space and applies a sparse autoencoder to uncover semantic deviations into interpretable factors.<n>Experiments across five benchmarks demonstrate that MISP-DPO consistently improves multimodal alignment over prior methods.
  arXiv  Detail & Related papers  (2025-09-30T03:24:09Z)
• Population-Proportional Preference Learning from Human Feedback: An Axiomatic Approach [6.578074497549894]
  We develop a novel preference learning framework capable of aligning aggregate opinions and policies proportionally with the true population distribution of evaluator preferences.<n>Our approach infers the feasible set of evaluator population distributions directly from pairwise comparison data.<n>We propose a soft-max relaxation method that smoothly trade-offs population-proportional representation with the selection of the Condorcet winner.
  arXiv  Detail & Related papers  (2025-06-05T22:15:07Z)
• In-context Ranking Preference Optimization [65.5489745857577]
  We propose an In-context Ranking Preference Optimization (IRPO) framework to optimize large language models (LLMs) based on ranking lists constructed during inference.<n>We show IRPO outperforms standard DPO approaches in ranking performance, highlighting its effectiveness in aligning LLMs with direct in-context ranking preferences.
  arXiv  Detail & Related papers  (2025-04-21T23:06:12Z)
• TIS-DPO: Token-level Importance Sampling for Direct Preference Optimization With Estimated Weights [73.9088920210495]
  We propose a token-level importance sampling DPO objective named TIS-DPO that assigns importance weights to each token based on its reward.<n>TIS-DPO significantly outperforms various baseline methods on harmlessness and helpfulness alignment and summarization tasks.
  arXiv  Detail & Related papers  (2024-10-06T04:03:00Z)
• 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)
• Pareto-Optimal Learning from Preferences with Hidden Context [17.590330740964266]
  We propose POPL, which enables pluralistic alignment by framing discrepant group preferences as objectives with potential trade-offs.<n>Our theoretical and empirical evaluations demonstrate that POPL surpasses baseline methods in learning sets of reward functions and policies.<n>We illustrate that POPL can also serve as a foundation for techniques optimizing specific notions of group fairness.
  arXiv  Detail & Related papers  (2024-06-21T18:57:38Z)
• Clipped-Objective Policy Gradients for Pessimistic Policy Optimization [3.2996723916635275]
  Policy gradient methods seek to produce monotonic improvement through bounded changes in policy outputs. In this work, we find that the performance of PPO, when applied to continuous action spaces, may be consistently improved through a simple change in objective. We show that the clipped-objective policy gradient (COPG) objective is on average "pessimistic" compared to both the PPO objective and (2) this pessimism promotes enhanced exploration.
  arXiv  Detail & Related papers  (2023-11-10T03:02:49Z)
• Well-being policy evaluation methodology based on WE pluralism [0.0]
  This study shifts from pluralism based on objective indicators to conceptual pluralism that emphasizes subjective context. By combining well-being and joint fact-finding on the narrow-wide WE consensus, the policy evaluation method is formulated.
  arXiv  Detail & Related papers  (2023-05-08T06:51:43Z)

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.