EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Parameter-Efficient Fine-Tuning of LLMs for Low-Power Sentiment Analysis Applications

Antariksh Jyotimilind Shreshthi, Sergio Spanò, Luca Di Nunzio, Riccardo La Cesa, Cristian Valenti Marco Re, Gian Carlo Cardarilli

JOIV : International Journal on Informatics Visualization · 30.01.2026 · 10 (1) · 232

DOI: 10.62527/joiv.10.1.4044 ↗

Kurz gefasst

Fachartikel in der Zeitschrift JOIV : International Journal on Informatics Visualization, online erschienen am 30.01.2026. Verfasst von Antariksh Jyotimilind Shreshthi und 5 weiteren Autor:innen. Die Arbeit wurde bisher 18-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper quantifies how parameter-efficient fine-tuning narrows the performance gap between compact transformers and massive Large Language Models (LLMs) on sentiment analysis while reducing memory, energy, and financial budgets. Experiments use the publicly released BERT-base-uncased checkpoint (12 layers, 110 M parameters), the Stanford Sentiment Treebank-2 benchmark (67,349 movie phrases with binary polarity labels), PyTorch and Hugging Face Transformers 4.41, Weights & Biases for experiment tracking, and an NVIDIA A100 GPU configured for mixed-precision computation. Low-Rank Adaptation (LoRA) adapters (rank 16) are injected into the query/key/value projections, so only 0.54 % of weights are updated. The model was trained for five epochs with AdamW, cosine scheduling, batch size 64, and optional 4-bit post-training quantization. Accuracy and macro-averaged F1 are logged every 25 steps. The LoRA-tuned model achieves 91% accuracy and 0.91 F1 on the SST-2 test set, an improvement of 40 percentage points over the off-the-shelf checkpoint and comparable to GPT-4o-mini (93%), while using fewer than 1⁄1500 of its parameters. Validation loss plateaued without overfitting, and 4-bit quantization compressed the model to 27 MB with

Zitieren

APA

Shreshthi, A. J., Spanò, S., Nunzio, L. D., Cesa, R. L., Re, C. V. M., & Cardarilli, G. C. (2026). Parameter-Efficient Fine-Tuning of LLMs for Low-Power Sentiment Analysis Applications. JOIV : International Journal on Informatics Visualization, 10(1), 232. https://doi.org/10.62527/joiv.10.1.4044

Vancouver

Shreshthi AJ, Spanò S, Nunzio LD, Cesa RL, Re CVM, Cardarilli GC. Parameter-Efficient Fine-Tuning of LLMs for Low-Power Sentiment Analysis Applications. JOIV : International Journal on Informatics Visualization. 2026;10(1):232. doi:10.62527/joiv.10.1.4044

BibTeX
@article{Shreshthi2026,
  author = {Shreshthi, Antariksh Jyotimilind and Spanò, Sergio and Nunzio, Luca Di and Cesa, Riccardo La and Re, Cristian Valenti Marco and Cardarilli, Gian Carlo},
  title = {Parameter-Efficient Fine-Tuning of LLMs for Low-Power Sentiment Analysis Applications},
  journal = {JOIV : International Journal on Informatics Visualization},
  year = {2026},
  volume = {10},
  number = {1},
  pages = {232},
  doi = {10.62527/joiv.10.1.4044}
}
Kostenlos starten →