GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO

TL;DR

GanitLLM is a Bengali mathematical reasoning model that uses a difficulty-aware curriculum training pipeline, significantly improving reasoning accuracy and Bengali token usage on math benchmarks.

Contribution

The paper introduces GanitLLM, a new Bengali math dataset with difficulty tags, and a curriculum-based training pipeline called Curriculum-GRPO for low-resource language reasoning.

Findings

GanitLLM-4B improves accuracy by +8 and +6 points on two benchmarks.

Increases Bengali reasoning tokens from 14% to over 88%.

Reduces solution length from 943 to 193 words.

Abstract

We present a Bengali mathematical reasoning model called GanitLLM (named after the Bangla word for mathematics, Ganit), together with a new difficulty-aware Bengali math corpus and a curriculum-based GRPO pipeline. Bengali is one of the world's most widely spoken languages, yet existing LLMs either reason in English and then translate, or simply fail on multi-step Bengali math, in part because reinforcement learning recipes are tuned for high-resource languages and collapse under reward sparsity in low-resource settings. To address this, we construct Ganit, a rigorously filtered and decontaminated Bengali math dataset with automatic difficulty tags derived from the pass@k of a strong evaluator model. Building on this dataset, we propose Curriculum-GRPO, which combines multi-stage training (SFT + GRPO) with difficulty-aware sampling and verifiable rewards for format, numerical correctness, and Bengali reasoning. On Bn-MGSM and Bn-MSVAMP, GanitLLM-4B improves over its Qwen3-4B base by +8 and +6 accuracy points, respectively, while increasing the percentage of Bengali reasoning tokens from 14% to over 88% and reducing average solution length from 943 to 193 words. Project page is available at https://dipta007.github.io/GanitLLM