Exploring the performance of ML model size for classification in relation to energy consumption
(2025) In Lecture Notes in Computer Science 16361. p.525-532- Abstract
- The use of large language models (LLMs) is being explored for a multitude of tasks in software engineering (SE), ranging from code generation to bug report assignment. Although LLMs provide impressive results, they require more time and energy than some other machine learning models. For some tasks, simpler models may be more sustainable than LLMs.
In this paper, we construct natural language classifiers of different complexity for a use case in the SE domain: commit message classification. We compare the performance of each model with the state-of-the-art with regard to energy consumption for training and inference. We find that simpler models based on Naïve Bayes and LSTM
perform similarly to LLMs, while using a fraction... (More) - The use of large language models (LLMs) is being explored for a multitude of tasks in software engineering (SE), ranging from code generation to bug report assignment. Although LLMs provide impressive results, they require more time and energy than some other machine learning models. For some tasks, simpler models may be more sustainable than LLMs.
In this paper, we construct natural language classifiers of different complexity for a use case in the SE domain: commit message classification. We compare the performance of each model with the state-of-the-art with regard to energy consumption for training and inference. We find that simpler models based on Naïve Bayes and LSTM
perform similarly to LLMs, while using a fraction of the energy, suggesting that choosing a small model can lead to significant reduction in power usage without compromising performance.
Replication package: https://doi.org/10.5281/zenodo.15641782 (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/8596bf17-0db4-4694-aafd-3ad33c1e49ed
- author
- Bexell, Andreas
LU
; Gullstrand Heander, Lo
LU
; Söderberg, Emma
LU
; Eldh, Sigrid
and Runeson, Per
LU
- organization
- publishing date
- 2025-11-20
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Product-focused software process improvement : 26th International Conference, PROFES 2025, Salerno, Italy, December 1–3, 2025 - 26th International Conference, PROFES 2025, Salerno, Italy, December 1–3, 2025
- series title
- Lecture Notes in Computer Science
- editor
- Scanniello, Giuseppe ; Lenarduzzi, Valentina ; Romano, Simone ; Vegas, Sira and Francese, Rita
- volume
- 16361
- pages
- 525 - 532
- publisher
- Springer
- external identifiers
-
- scopus:105023309214
- ISSN
- 1611-3349
- ISBN
- 978-3-032-12089-2
- 978-3-032-12088-5
- DOI
- 10.1007/978-3-032-12089-2_38
- language
- English
- LU publication?
- yes
- id
- 8596bf17-0db4-4694-aafd-3ad33c1e49ed
- date added to LUP
- 2025-11-25 10:44:29
- date last changed
- 2025-12-05 04:00:09
@inproceedings{8596bf17-0db4-4694-aafd-3ad33c1e49ed,
abstract = {{The use of large language models (LLMs) is being explored for a multitude of tasks in software engineering (SE), ranging from code generation to bug report assignment. Although LLMs provide impressive results, they require more time and energy than some other machine learning models. For some tasks, simpler models may be more sustainable than LLMs. <br/><br/>In this paper, we construct natural language classifiers of different complexity for a use case in the SE domain: commit message classification. We compare the performance of each model with the state-of-the-art with regard to energy consumption for training and inference. We find that simpler models based on Naïve Bayes and LSTM <br/>perform similarly to LLMs, while using a fraction of the energy, suggesting that choosing a small model can lead to significant reduction in power usage without compromising performance.<br/><br/>Replication package: https://doi.org/10.5281/zenodo.15641782}},
author = {{Bexell, Andreas and Gullstrand Heander, Lo and Söderberg, Emma and Eldh, Sigrid and Runeson, Per}},
booktitle = {{Product-focused software process improvement : 26th International Conference, PROFES 2025, Salerno, Italy, December 1–3, 2025}},
editor = {{Scanniello, Giuseppe and Lenarduzzi, Valentina and Romano, Simone and Vegas, Sira and Francese, Rita}},
isbn = {{978-3-032-12089-2}},
issn = {{1611-3349}},
language = {{eng}},
month = {{11}},
pages = {{525--532}},
publisher = {{Springer}},
series = {{Lecture Notes in Computer Science}},
title = {{Exploring the performance of ML model size for classification in relation to energy consumption}},
url = {{http://dx.doi.org/10.1007/978-3-032-12089-2_38}},
doi = {{10.1007/978-3-032-12089-2_38}},
volume = {{16361}},
year = {{2025}},
}