Riazi-8B: An Urdu Large Language Model for Mathematical Reasoning

2026-06-24

AI Summary

This work addresses the scarcity of mathematical reasoning datasets and adapted models for low-resource languages such as Urdu, which severely limits the performance of large language models on multi-step math problems in such languages. The authors propose a two-stage adaptation strategy: first conducting continued pretraining on Urdu Wikipedia, followed by supervised fine-tuning using a Chain-of-Thought dataset constructed by translating GSM8K into Urdu. This approach uniquely integrates language adaptation with reasoning-oriented fine-tuning, effectively transferring mathematical reasoning capabilities to Urdu. Experimental results on the MGSM-Urdu benchmark demonstrate that the proposed method significantly outperforms existing Urdu instruction-tuned models, achieving consistent improvements in answer accuracy, reasoning quality, response completeness, and language generation fluency, thereby filling critical gaps in both data and modeling for mathematical reasoning in Urdu.

Abstract

Recent LLMs demonstrate strong mathematical reasoning capabilities, but existing gains rely heavily on English-centric training resources and benchmarks. As a result, reasoning performance degrades substantially in low-resource languages such as Urdu, where reasoning-oriented datasets and adapted models remain scarce. Urdu lacks both reasoning-oriented resources and models adapted for multi-step mathematical problem solving, limiting the applicability of recent progress to Urdu-speaking users. We address this gap through Riazi-8B, an Urdu mathematical reasoning model developed through a two-step adaptation process comprising continued pre-training on Urdu Wikipedia and supervised fine-tuning on Urdu Chain-of-Thought data derived from GSM8K. We evaluate Riazi-8B on MGSM-Urdu against existing Urdu instruction-tuned models. Our results show consistent improvements in answer correctness, reasoning quality, response completeness, and Urdu generation. Our findings demonstrate that combining Urdu language adaptation with reasoning-focused fine-tuning is an effective strategy for extending mathematical reasoning capabilities to low-resource languages.

Problem

• mathematical reasoning
• low-resource languages
• Urdu
• large language models
• reasoning datasets

Innovation

• Urdu LLM
• mathematical reasoning
• Chain-of-Thought
• low-resource languages
• two-stage adaptation