BhashaKritika: Building Synthetic Pretraining Data at Scale for Indic Languages
This addresses the uneven distribution of LLM benefits across languages, particularly for Indic language communities, though it represents an incremental improvement in synthetic data generation methods.
The researchers tackled the problem of generating high-quality synthetic pretraining data for low-resource Indic languages by constructing BhashaKritika, a 540B token dataset using 5 techniques across 10 languages, and introduced a modular evaluation pipeline for quality control.
In the context of pretraining of Large Language Models (LLMs), synthetic data has emerged as an alternative for generating high-quality pretraining data at scale. This is particularly beneficial in low-resource language settings where the benefits of recent LLMs have been unevenly distributed across languages. In this work, we present a systematic study on the generation and evaluation of synthetic multilingual pretraining data for Indic languages, where we construct a large-scale synthetic dataset BhashaKritika, comprising 540B tokens using 5 different techniques for 10 languages. We explore the impact of grounding generation in documents, personas, and topics. We analyze how language choice, both in the prompt instructions and document grounding, affects data quality, and we compare translations of English content with native generation in Indic languages. To support scalable and language-sensitive evaluation, we introduce a modular quality evaluation pipeline that integrates script and language detection, metadata consistency checks, n-gram repetition analysis, and perplexity-based filtering using KenLM models. Our framework enables robust quality control across diverse scripts and linguistic contexts. Empirical results through model runs reveal key trade-offs in generation strategies and highlight best practices for constructing effective multilingual corpora.