KAZ-Div: A Model-Agnostic Framework for Quality-Constrained Diversity in Kazakh Language Generation
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.
Authors
- Altynbek Seitenov (ORCID: https://orcid.org/0000-0001-5777-4363)
- Pınar Sarısaray Bölük (ORCID: https://orcid.org/0000-0001-8274-8423)
- Zhansaya Makhambetova (ORCID: https://orcid.org/0000-0001-5024-0289)
- Arailym Tleubayeva (ORCID: https://orcid.org/0000-0001-9560-9756)
Institutions
- University of Leeds (GB)
- Istanbul University-Cerrahpaşa (TR)
- Astana IT University (KZ)
- Istanbul University (TR)
Publication Details
- Journal
- Big Data and Cognitive Computing
- Published
- 2026-11-16
- DOI
- https://doi.org/10.3390/bdcc10110391
- Primary Topic
- Topic Modeling
- Type
- article
- Field-Weighted Citation Impact
- 0.00