EVOLUTION OF LARGE LANGUAGE MODELS: A COMPREHENSIVE REVIEW OF VERSIONS AND ADVANCEMENTS
Al-Shodhana · 30.01.2026 · 14 (1) · 501-521
DOI: 10.70644/as.v14.i1.41 ↗Kurz gefasst
Fachartikel in der Zeitschrift Al-Shodhana, online erschienen am 30.01.2026. Verfasst von Ms A Vanaja und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
This paper reviews the evolution of large language models from early recurrent networks to transformer-centred systems. We ask which architectural changes matter most, how training data and scale shape capabilities, and what computational trade-offs arise. Through a focused systematic review we categorised model families, compared dataset and compute scales, and synthesised optimisation strategies. Findings show transformer improvements—especially self-attention and long-context handling—drive contextual gains, while diverse, well-curated corpora underpin emergent abilities. Scaling increases compute and energy demands, and persistent issues remain in eficiency, bias and reproducibility. We recommend standardised benchmarks, stronger dataset quality controls and wider adoption of resource-aware techniques such as quantisation, pruning and parameter-eficient fine-tuning. These measures support balanced model selection and responsible deployment in sensitive domains and governance.
Zitieren
APA
Vanaja, M. A., Sumanth, M. R., Noronha, V. B., & Dsouza, R. R. (2026). EVOLUTION OF LARGE LANGUAGE MODELS: A COMPREHENSIVE REVIEW OF VERSIONS AND ADVANCEMENTS. Al-Shodhana, 14(1), 501-521. https://doi.org/10.70644/as.v14.i1.41
Vancouver
Vanaja MA, Sumanth MR, Noronha VB, Dsouza RR. EVOLUTION OF LARGE LANGUAGE MODELS: A COMPREHENSIVE REVIEW OF VERSIONS AND ADVANCEMENTS. Al-Shodhana. 2026;14(1):501-521. doi:10.70644/as.v14.i1.41
BibTeX
@article{Vanaja2026,
author = {Vanaja, Ms A and Sumanth, M R and Noronha, Vian Branson and Dsouza, Ronsy Ronald},
title = {EVOLUTION OF LARGE LANGUAGE MODELS: A COMPREHENSIVE REVIEW OF VERSIONS AND ADVANCEMENTS},
journal = {Al-Shodhana},
year = {2026},
volume = {14},
number = {1},
pages = {501--521},
doi = {10.70644/as.v14.i1.41}
}
EUVIMED