Relic: Enhancing Reward Model Generalization for Low-Resource Indic Languages with Few-Shot Examples

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ghosal, Soumya Suvra, Singh, Vaibhav, Ghosh, Akash, Pal, Soumyabrata, Baidya, Subhadip, Saha, Sriparna, Manocha, Dinesh
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918065721049088
author Ghosal, Soumya Suvra
Singh, Vaibhav
Ghosh, Akash
Pal, Soumyabrata
Baidya, Subhadip
Saha, Sriparna
Manocha, Dinesh
author_facet Ghosal, Soumya Suvra
Singh, Vaibhav
Ghosh, Akash
Pal, Soumyabrata
Baidya, Subhadip
Saha, Sriparna
Manocha, Dinesh
contents Reward models are essential for aligning large language models (LLMs) with human preferences. However, most open-source multilingual reward models are primarily trained on preference datasets in high-resource languages, resulting in unreliable reward signals for low-resource Indic languages. Collecting large-scale, high-quality preference data for these languages is prohibitively expensive, making preference-based training approaches impractical. To address this challenge, we propose RELIC, a novel in-context learning framework for reward modeling in low-resource Indic languages. RELIC trains a retriever with a pairwise ranking objective to select in-context examples from auxiliary high-resource languages that most effectively highlight the distinction between preferred and less-preferred responses. Extensive experiments on three preference datasets- PKU-SafeRLHF, WebGPT, and HH-RLHF-using state-of-the-art open-source reward models demonstrate that RELIC significantly improves reward model accuracy for low-resource Indic languages, consistently outperforming existing example selection methods. For example, on Bodo-a low-resource Indic language-using a LLaMA-3.2-3B reward model, RELIC achieves a 12.81% and 10.13% improvement in accuracy over zero-shot prompting and state-of-the-art example selection method, respectively.
format Preprint
id arxiv_https___arxiv_org_abs_2506_16502
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Relic: Enhancing Reward Model Generalization for Low-Resource Indic Languages with Few-Shot Examples
Ghosal, Soumya Suvra
Singh, Vaibhav
Ghosh, Akash
Pal, Soumyabrata
Baidya, Subhadip
Saha, Sriparna
Manocha, Dinesh
Computation and Language
Artificial Intelligence
Reward models are essential for aligning large language models (LLMs) with human preferences. However, most open-source multilingual reward models are primarily trained on preference datasets in high-resource languages, resulting in unreliable reward signals for low-resource Indic languages. Collecting large-scale, high-quality preference data for these languages is prohibitively expensive, making preference-based training approaches impractical. To address this challenge, we propose RELIC, a novel in-context learning framework for reward modeling in low-resource Indic languages. RELIC trains a retriever with a pairwise ranking objective to select in-context examples from auxiliary high-resource languages that most effectively highlight the distinction between preferred and less-preferred responses. Extensive experiments on three preference datasets- PKU-SafeRLHF, WebGPT, and HH-RLHF-using state-of-the-art open-source reward models demonstrate that RELIC significantly improves reward model accuracy for low-resource Indic languages, consistently outperforming existing example selection methods. For example, on Bodo-a low-resource Indic language-using a LLaMA-3.2-3B reward model, RELIC achieves a 12.81% and 10.13% improvement in accuracy over zero-shot prompting and state-of-the-art example selection method, respectively.
title Relic: Enhancing Reward Model Generalization for Low-Resource Indic Languages with Few-Shot Examples
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2506.16502