Saved in:
Bibliographic Details
Main Authors: Bhattacharjee, Payel, Simeone, Osvaldo, Tandon, Ravi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.17658
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911715034136576
author Bhattacharjee, Payel
Simeone, Osvaldo
Tandon, Ravi
author_facet Bhattacharjee, Payel
Simeone, Osvaldo
Tandon, Ravi
contents Reward modeling is central to alignment pipelines such as RLHF, RLAIF, and PPO-based policy optimization, yet its reliability is constrained by limited and heterogeneous human preference data that are expensive to collect at scale. While synthetic augmentation can expand preference supervision, existing methods often augment uniformly or at the representation level, without targeting examples where the reward model is uncertain or prone to mis-ranking. In this paper, we introduce MARS (Margin and Semantic-Aware Data Augmentation for Reward Modeling), an adaptive augmentation framework that prioritizes low-margin preference pairs and uses semantic distance as a second layer for refinement to enhance the contrast between the chosen and rejected responses. Across multiple preference datasets, reward-model backbones, downstream alignment settings, and benchmarks including RewardBench and AlpacaEval, MARS improves both reward-model quality and alignment performance over existing baselines. Our results show that reward-model augmentation is most effective when guided by both model margins and semantic structure.
format Preprint
id arxiv_https___arxiv_org_abs_2602_17658
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MARS: Margin and Semantic-Aware Data Augmentation for Reward Modeling
Bhattacharjee, Payel
Simeone, Osvaldo
Tandon, Ravi
Machine Learning
Artificial Intelligence
Information Theory
Reward modeling is central to alignment pipelines such as RLHF, RLAIF, and PPO-based policy optimization, yet its reliability is constrained by limited and heterogeneous human preference data that are expensive to collect at scale. While synthetic augmentation can expand preference supervision, existing methods often augment uniformly or at the representation level, without targeting examples where the reward model is uncertain or prone to mis-ranking. In this paper, we introduce MARS (Margin and Semantic-Aware Data Augmentation for Reward Modeling), an adaptive augmentation framework that prioritizes low-margin preference pairs and uses semantic distance as a second layer for refinement to enhance the contrast between the chosen and rejected responses. Across multiple preference datasets, reward-model backbones, downstream alignment settings, and benchmarks including RewardBench and AlpacaEval, MARS improves both reward-model quality and alignment performance over existing baselines. Our results show that reward-model augmentation is most effective when guided by both model margins and semantic structure.
title MARS: Margin and Semantic-Aware Data Augmentation for Reward Modeling
topic Machine Learning
Artificial Intelligence
Information Theory
url https://arxiv.org/abs/2602.17658