KAZ-Div: A Model-Agnostic Framework for Quality-Constrained Diversity in Kazakh Language Generation
Big Data and Cognitive Computing · 16.11.2026 · 10 (11) · 391
DOI: 10.3390/bdcc10110391 ↗Kurz gefasst
Fachartikel in der Zeitschrift Big Data and Cognitive Computing, online erschienen am 16.11.2026. Verfasst von Arailym Tleubayeva und 3 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan.
Abstract (Originalsprache)
Large language models (LLMs) often produce repetitive outputs, limiting linguistic diversity in lower-resource languages. This study introduces KAZ-Div, a model-agnostic inference-time framework combining multi-candidate generation, task-specific validity filtering, semantic redundancy management, and cluster-based response selection. We evaluated four LLMs across three Kazakh-language tasks: paraphrase generation, grounded question generation, and synthetic intent-data generation. Evaluation included lexical and embedding-based diversity metrics, automated task-specific constraint checks, stage-wise ablation, alternative prompting and decoding baselines, independent embedding-based fidelity validation, parameter-sensitivity analysis, and downstream intent classification. Compared with conventional single-response generation, KAZ-Div generally increased lexical diversity; for paraphrasing, Distinct-2 increased from 0.322 to 0.719 for GPT-5.6 Sol and from 0.559 to 0.840 for Claude Sonnet 4.6. However, embedding-based semantic-diversity and source-fidelity effects varied across models. Same-pool ablation revealed no consistent lexical-diversity advantage over random candidate selection, indicating that multi-candidate generation accounted for much of the observed improvement. Independent evaluations using LaBSE and multilingual-e5-base supported the principal source-fidelity patterns but did not establish an additional fidelity benefit from semantic curation. The current experiments do not demonstrate a general downstream benefit of KAZ-Div: after matching the number of unique training examples, the differences for GPT-5.6 Sol and Claude Sonnet 4.6 were not statistically significant, and multi-seed XLM-RoBERTa effects varied in direction across generators. Grammaticality, naturalness, task validity, and syntactic variation were not independently assessed. Overall, KAZ-Div provides a structured framework for quality-constrained candidate management, whose practical value depends on the generator, task-specific requirements, and computational budget.
Förderung
- Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan
Zitieren
APA
Tleubayeva, A., Makhambetova, Z., Seitenov, A., & Boluk, P. S. (2026). KAZ-Div: A Model-Agnostic Framework for Quality-Constrained Diversity in Kazakh Language Generation. Big Data and Cognitive Computing, 10(11), 391. https://doi.org/10.3390/bdcc10110391
Vancouver
Tleubayeva A, Makhambetova Z, Seitenov A, Boluk PS. KAZ-Div: A Model-Agnostic Framework for Quality-Constrained Diversity in Kazakh Language Generation. Big Data and Cognitive Computing. 2026;10(11):391. doi:10.3390/bdcc10110391
BibTeX
@article{Tleubayeva2026,
author = {Tleubayeva, Arailym and Makhambetova, Zhansaya and Seitenov, Altynbek and Boluk, Pinar Sarisaray},
title = {KAZ-Div: A Model-Agnostic Framework for Quality-Constrained Diversity in Kazakh Language Generation},
journal = {Big Data and Cognitive Computing},
year = {2026},
volume = {10},
number = {11},
pages = {391},
doi = {10.3390/bdcc10110391}
}
EUVIMED