R3: Robust Rubric-Agnostic Reward Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Anugraha, David, Tang, Zilu, Miranda, Lester James V., Zhao, Hanyang, Farhansyah, Mohammad Rifqi, Kuwanto, Garry, Wijaya, Derry, Winata, Genta Indra
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909798099845120
author Anugraha, David
Tang, Zilu
Miranda, Lester James V.
Zhao, Hanyang
Farhansyah, Mohammad Rifqi
Kuwanto, Garry
Wijaya, Derry
Winata, Genta Indra
author_facet Anugraha, David
Tang, Zilu
Miranda, Lester James V.
Zhao, Hanyang
Farhansyah, Mohammad Rifqi
Kuwanto, Garry
Wijaya, Derry
Winata, Genta Indra
contents Reward models are essential for aligning language model outputs with human preferences, yet existing approaches often lack both controllability and interpretability. These models are typically optimized for narrow objectives, limiting their generalizability to broader downstream tasks. Moreover, their scalar outputs are difficult to interpret without contextual reasoning. To address these limitations, we introduce $\shortmethodname$, a novel reward modeling framework that is rubric-agnostic, generalizable across evaluation dimensions, and provides interpretable, reasoned score assignments. $\shortmethodname$ enables more transparent and flexible evaluation of language models, supporting robust alignment with diverse human values and use cases. Our models, data, and code are available as open source at https://github.com/rubricreward/r3.
format Preprint
id arxiv_https___arxiv_org_abs_2505_13388
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle R3: Robust Rubric-Agnostic Reward Models
Anugraha, David
Tang, Zilu
Miranda, Lester James V.
Zhao, Hanyang
Farhansyah, Mohammad Rifqi
Kuwanto, Garry
Wijaya, Derry
Winata, Genta Indra
Computation and Language
Artificial Intelligence
Machine Learning
Reward models are essential for aligning language model outputs with human preferences, yet existing approaches often lack both controllability and interpretability. These models are typically optimized for narrow objectives, limiting their generalizability to broader downstream tasks. Moreover, their scalar outputs are difficult to interpret without contextual reasoning. To address these limitations, we introduce $\shortmethodname$, a novel reward modeling framework that is rubric-agnostic, generalizable across evaluation dimensions, and provides interpretable, reasoned score assignments. $\shortmethodname$ enables more transparent and flexible evaluation of language models, supporting robust alignment with diverse human values and use cases. Our models, data, and code are available as open source at https://github.com/rubricreward/r3.
title R3: Robust Rubric-Agnostic Reward Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2505.13388