EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Research on Multi-Task Natural Language Understanding and Generation Methods Based on Large Language Models

Eleanor Rigby, Thomas Anderson, Julianne Moore

Global Media and Social Sciences Research Journal · 05.04.2026 · 7 (2) · 15-24

DOI: 10.71465/gmssrj182 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Global Media and Social Sciences Research Journal, online erschienen am 05.04.2026. Verfasst von Eleanor Rigby und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid evolution of large language models has fundamentally transformed the landscape of computational linguistics, enabling unprecedented capabilities across a diverse array of tasks. While historically natural language understanding and natural language generation were treated as disparate domains requiring specialized architectures, contemporary paradigms leverage unified models to concurrently execute classification, extraction, summarization, and open-ended text creation. This paper presents a comprehensive investigation into multi-task methods based on large language models, aiming to bridge the operational gap between understanding and generation objectives. We propose a novel algorithmic framework that mitigates negative transfer and gradient interference, phenomena frequently observed when optimizing models for competing downstream tasks. By introducing a dynamic task-weighting mechanism and a unified instruction-tuned formatting approach, our methodology aligns generative sequences with discriminative representation learning. We present extensive empirical analyses across established benchmarks, demonstrating that our approach significantly enhances parameter efficiency while achieving superior or competitive performance compared to single-task models. The findings illuminate the underlying mechanisms of shared representation spaces and provide critical insights for the deployment of versatile, multi-capable computational systems in complex real-world environments.

Zitieren

APA

Rigby, E., Anderson, T., & Moore, J. (2026). Research on Multi-Task Natural Language Understanding and Generation Methods Based on Large Language Models. Global Media and Social Sciences Research Journal, 7(2), 15-24. https://doi.org/10.71465/gmssrj182

Vancouver

Rigby E, Anderson T, Moore J. Research on Multi-Task Natural Language Understanding and Generation Methods Based on Large Language Models. Global Media and Social Sciences Research Journal. 2026;7(2):15-24. doi:10.71465/gmssrj182

BibTeX
@article{Rigby2026,
  author = {Rigby, Eleanor and Anderson, Thomas and Moore, Julianne},
  title = {Research on Multi-Task Natural Language Understanding and Generation Methods Based on Large Language Models},
  journal = {Global Media and Social Sciences Research Journal},
  year = {2026},
  volume = {7},
  number = {2},
  pages = {15--24},
  doi = {10.71465/gmssrj182}
}
Kostenlos starten →