BhashaKritika: Building Synthetic Pretraining Data at Scale for Indic Languages

• URL: http://arxiv.org/abs/2511.10338v2
• Date: Sun, 16 Nov 2025 13:08:22 GMT
• Title: BhashaKritika: Building Synthetic Pretraining Data at Scale for Indic Languages
• Authors: Guduru Manoj, Neel Prabhanjan Rachamalla, Ashish Kulkarni, Gautam Rajeev, Jay Piplodiya, Arul Menezes, Shaharukh Khan, Souvik Rana, Manya Sah, Chandra Khatri, Shubham Agarwal,
• Abstract summary: We present a systematic study on the generation and evaluation of synthetic multilingual pretraining data for Indic languages.<n>We construct a large-scale synthetic dataset BhashaKritika, comprising 540B tokens using 5 different techniques for 10 languages.<n>We analyze how language choice, both in the prompt instructions and document grounding, affects data quality.
• Score: 4.279942349440352
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• What Language is This? Ask Your Tokenizer [32.28976119949841]
  Language Identification (LID) is an important component of many multilingual natural language processing pipelines.<n>We introduce UniLID, a simple and efficient LID method based on the UnigramLM tokenization algorithm.<n>Our formulation is data- and compute-efficient, supports incremental addition of new languages without retraining existing models.
  arXiv  Detail & Related papers  (2026-02-19T18:58:39Z)
• Fluent Alignment with Disfluent Judges: Post-training for Lower-resource Languages [16.671158083515373]
  We develop a fluent preference-aligned language model without instruction-tuning data in the target language.<n>Our approach uses an on-policy training method, which we compare with two common approaches.<n>We conduct a case study on Norwegian Bokml and evaluate fluency through native-speaker assessments.
  arXiv  Detail & Related papers  (2025-12-09T16:31:48Z)
• LangGPS: Language Separability Guided Data Pre-Selection for Joint Multilingual Instruction Tuning [49.22807995935406]
  Joint multilingual instruction tuning is a widely adopted approach to improve the multilingual instruction-following ability and downstream performance of large language models (LLMs)<n>Existing selection methods, often based on features like text quality, diversity, or task relevance, typically overlook the intrinsic linguistic structure of multilingual data.<n>We propose LangGPS, a lightweight two-stage pre-selection framework guided by language separability which quantifies how well samples in different languages can be distinguished in the model's representation space.
  arXiv  Detail & Related papers  (2025-11-13T12:02:32Z)
• The role of synthetic data in Multilingual, Multi-cultural AI systems: Lessons from Indic Languages [18.087937520281965]
  We introduce Updesh, a large-scale synthetic instruction-following dataset comprising 9.5M data points across 13 Indian languages.<n>A comprehensive evaluation incorporating both automated metrics and human annotation across 10k assessments indicates that generated data is high quality.<n>Models trained on Updesh consistently achieve significant gains on generative tasks and remain competitive on multiple-choice style NLU tasks.
  arXiv  Detail & Related papers  (2025-09-25T15:13:00Z)
• Judging Quality Across Languages: A Multilingual Approach to Pretraining Data Filtering with Language Models [52.22235443948351]
  High-quality multilingual training data is essential for effectively pretraining large language models (LLMs)<n>Here, we introduce JQL, a systematic approach that efficiently curates diverse and high-quality multilingual data at scale.<n>JQL distills LLMs' annotation capabilities into lightweight annotators based on pretrained multilingual embeddings.
  arXiv  Detail & Related papers  (2025-05-28T11:06:54Z)
• Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training [58.696660064190475]
  We find that the existence of code-switching, alternating between different languages within a context, is key to multilingual capabilities.<n>To better explore the power of code-switching for language alignment during pre-training, we investigate the strategy of synthetic code-switching.
  arXiv  Detail & Related papers  (2025-04-02T15:09:58Z)
• Enhancing Multilingual Language Models for Code-Switched Input Data [0.0]
  This research investigates if pre-training Multilingual BERT (mBERT) on code-switched datasets improves the model's performance on critical NLP tasks.<n>We use a dataset of Spanglish tweets for pre-training and evaluate the pre-trained model against a baseline model.<n>Our findings show that our pre-trained mBERT model outperforms or matches the baseline model in the given tasks, with the most significant improvements seen for parts of speech tagging.
  arXiv  Detail & Related papers  (2025-03-11T02:49:41Z)
• Learning Phonotactics from Linguistic Informants [54.086544221761486]
  Our model iteratively selects or synthesizes a data-point according to one of a range of information-theoretic policies. We find that the information-theoretic policies that our model uses to select items to query the informant achieve sample efficiency comparable to, or greater than, fully supervised approaches.
  arXiv  Detail & Related papers  (2024-05-08T00:18:56Z)
• GradSim: Gradient-Based Language Grouping for Effective Multilingual Training [13.730907708289331]
  We propose GradSim, a language grouping method based on gradient similarity. Our experiments on three diverse multilingual benchmark datasets show that it leads to the largest performance gains. Besides linguistic features, the topics of the datasets play an important role for language grouping.
  arXiv  Detail & Related papers  (2023-10-23T18:13:37Z)
• How do languages influence each other? Studying cross-lingual data sharing during LM fine-tuning [14.02101305717738]
  Multilingual large language models (MLLMs) are jointly trained on data from many different languages. It remains unclear to what extent, and under which conditions, languages rely on each other's data. We find that MLLMs rely on data from multiple languages from the early stages of fine-tuning and that this reliance gradually increases as fine-tuning progresses.
  arXiv  Detail & Related papers  (2023-05-22T17:47:41Z)
• Multi-lingual Evaluation of Code Generation Models [82.7357812992118]
  We present new benchmarks on evaluation code generation models: MBXP and Multilingual HumanEval, and MathQA-X. These datasets cover over 10 programming languages. We are able to assess the performance of code generation models in a multi-lingual fashion.
  arXiv  Detail & Related papers  (2022-10-26T17:17:06Z)
• UNKs Everywhere: Adapting Multilingual Language Models to New Scripts [103.79021395138423]
  Massively multilingual language models such as multilingual BERT (mBERT) and XLM-R offer state-of-the-art cross-lingual transfer performance on a range of NLP tasks. Due to their limited capacity and large differences in pretraining data, there is a profound performance gap between resource-rich and resource-poor target languages. We propose novel data-efficient methods that enable quick and effective adaptation of pretrained multilingual models to such low-resource languages and unseen scripts.
  arXiv  Detail & Related papers  (2020-12-31T11:37:28Z)
• Comparison of Interactive Knowledge Base Spelling Correction Models for Low-Resource Languages [81.90356787324481]
  Spelling normalization for low resource languages is a challenging task because the patterns are hard to predict. This work shows a comparison of a neural model and character language models with varying amounts on target language data. Our usage scenario is interactive correction with nearly zero amounts of training examples, improving models as more data is collected.
  arXiv  Detail & Related papers  (2020-10-20T17:31:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.