RMGAP: Benchmarking the Generalization of Reward Models across Diverse Preferences

• URL: http://arxiv.org/abs/2605.01831v1
• Date: Sun, 03 May 2026 11:45:08 GMT
• Title: RMGAP: Benchmarking the Generalization of Reward Models across Diverse Preferences
• Abstract summary: This work focuses on how to evaluate the generalizability of reward models.<n>We introduce RMGAP, a benchmark comprising 1,097 instances across Chat, Writing, Reasoning, and Safety domains.<n>Even the best RM achieves only 49.27% Best-of-N accuracy, highlighting considerable room for improvement in reward model generalization.
• Score: 3.498051445282382
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reinforcement Learning from Human Feedback has become the standard paradigm for language model alignment, where reward models directly determine alignment effectiveness. In this work, we focus on how to evaluate the generalizability of reward models. By "generalizability", we mean the ability of RMs to correctly rank responses to align with diverse user preferences. However, existing reward model benchmarks are typically designed around a universal preference, failing to assess this generalization. To address this critical gap, we introduce RMGAP, a benchmark comprising 1,097 instances across Chat, Writing, Reasoning, and Safety domains. Since different users exhibit diverse preferences for the same task, we first generate four distinct responses with different linguistic profiles for each collected prompt. However, the original prompt set lacks the specificity to convey different preferences. We therefore construct tailored prompts by contrasting these candidates and designing scenarios in which one response becomes the uniquely appropriate choice. Moreover, we observe that users often express the same preference using different phrasings, and thus extend each prompt with two paraphrased variants. Our evaluation of 24 state-of-the-art RMs reveals their substantial limitations: even the best RM achieves only 49.27% Best-of-N accuracy, highlighting considerable room for improvement in reward model generalization. Related data and code are available at https://github.com/nanzhi84/RMGAP.

Related papers

• VIBE-Bench: Evaluating Personalized Large Language Models When Profiles Don't Mean Preferences [40.195544693657396]
  Existing benchmarks largely assume that preference can be retrieved from semantically related history.<n>We study an underexplored but practically important regime, profile-preference conceptual misalignment.<n>We introduce VIBE-Bench, a benchmark with two psychology-grounded tasks and 12,239 dialogues.
  arXiv  Detail & Related papers  (2026-09-01T08:45:28Z)
• A Consensus-Based Framework for Relative Preference Evaluation of Large Language Models [0.0]
  This paper introduces a consensus-based evaluation framework that measures relative preference among model-generated responses rather than absolute correctness.<n>We conduct a controlled study using five state-of-the-art LLMs across multiple domains, including programming, general knowledge, safety, logical reasoning, and mathematics.<n>Scores are aggregated into a Relative Intelligence Index (RII), representing how frequently a model's responses are preferred by other models.
  arXiv  Detail & Related papers  (2026-07-19T08:59:19Z)
• Response Time Enhances Alignment with Heterogeneous Preferences [49.69696266152175]
  We show that augmenting preference datasets with a simple, secondary signal can restore identifiability of the population's average preference.<n>Our results bring promises and open up new opportunities for future data-collection pipelines.
  arXiv  Detail & Related papers  (2026-05-07T22:05:23Z)
• No Single Best Model for Diversity: Learning a Router for Sample Diversity [69.53166985556759]
  In this paper, we study methods to elicit a comprehensive set of valid responses.<n>We introduce textbfdiversity coverage, a metric that measures the total quality scores assigned to each textbfunique answer.<n>Per each prompt, there exists a model that outperforms all other models significantly at generating a diverse answer set.
  arXiv  Detail & Related papers  (2026-04-02T17:58:37Z)
• P-GenRM: Personalized Generative Reward Model with Test-time User-based Scaling [66.55381105691818]
  We propose P-GenRM, the first Personalized Generative Reward Model with test-time user-based scaling.<n>P-GenRM transforms preference signals into structured evaluation chains that derive adaptive personas and scoring rubrics.<n>It further clusters users into User Prototypes and introduces a dual-granularity scaling mechanism.
  arXiv  Detail & Related papers  (2026-02-12T16:07:22Z)
• Robust Preference Alignment via Directional Neighborhood Consensus [13.313830197011983]
  We introduce Robust Preference Selection (RPS), a post-hoc, training-free method by leveraging directional neighborhood consensus.<n>RPS samples multiple responses from a local neighborhood of related preferences to create a superior candidate pool.<n>Our work presents a practical, theoretically-grounded solution for enhancing the reliability of preference-aligned models.
  arXiv  Detail & Related papers  (2025-10-23T12:39:20Z)
• Reinforcing Video Reasoning with Focused Thinking [65.85683941058916]
  We propose TW-GRPO, a novel framework that enhances visual reasoning with focused thinking and dense reward granularity.<n>Specifically, we employ a token weighting mechanism that prioritizes tokens with high informational density.<n>We also reformulate RL training by shifting from single-choice to multi-choice QA tasks.
  arXiv  Detail & Related papers  (2025-05-30T15:42:19Z)
• Data-adaptive Safety Rules for Training Reward Models [9.172858508920696]
  Reinforcement Learning from Human Feedback (RLHF) is commonly employed to tailor models to human preferences.<n>In this paper, we propose a dynamic method that adaptively selects the most important rules for each response pair.<n>As of January 25, 2025, our model achieved the highest safety performance on the leaderboard, surpassing various larger models.
  arXiv  Detail & Related papers  (2025-01-26T08:49:46Z)
• Beyond the Binary: Capturing Diverse Preferences With Reward Regularization [15.518838657050173]
  We argue that this reliance on binary choices does not capture the broader, aggregate preferences of the target user in real-world tasks.<n>We introduce a simple yet effective method that augments existing binary preference datasets with synthetic preference judgments to estimate potential user disagreement.
  arXiv  Detail & Related papers  (2024-12-05T02:35:46Z)
• ComPO: Community Preferences for Language Model Personalization [122.54846260663922]
  ComPO is a method to personalize preference optimization in language models. We collect and release ComPRed, a question answering dataset with community-level preferences from Reddit.
  arXiv  Detail & Related papers  (2024-10-21T14:02:40Z)
• 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)
• Getting MoRE out of Mixture of Language Model Reasoning Experts [71.61176122960464]
  We propose a Mixture-of-Reasoning-Experts (MoRE) framework that ensembles diverse specialized language models. We specialize the backbone language model with prompts optimized for different reasoning categories, including factual, multihop, mathematical, and commonsense reasoning. Our human study confirms that presenting expert predictions and the answer selection process helps annotators more accurately calibrate when to trust the system's output.
  arXiv  Detail & Related papers  (2023-05-24T02:00:51Z)

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.