Is Generative AI Hiding in Your Phone? Fine-tuning SLMs to Generate JSON-Adherent Output
(2026) EDAN70 20261Department of Computer Science
- Abstract
- How long will it be before our modern society is wrapped in generative AI without our discernible knowledge? What happens when the line between real and synthetic is practically non-existent? This study explores how a Qwen3-0.6B model, designed for mobile and edge deployment, can be fine-tuned to generate JSON-schematic data for workout and fitness. It shows that even smaller models are well-enough to design parsable output, ready to be used by an application - thus hiding its AI trademarks. In the future, studies on even smaller models may be appropriate, as well as discussions regarding its ethics.
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9249673
- author
- Angenius, Simon LU
- supervisor
- organization
- course
- EDAN70 20261
- year
- 2026
- type
- L3 - Miscellaneous, Projetcs etc.
- subject
- language
- English
- id
- 9249673
- date added to LUP
- 2026-09-01 13:57:52
- date last changed
- 2026-09-01 13:57:52
@misc{9249673,
abstract = {{How long will it be before our modern society is wrapped in generative AI without our discernible knowledge? What happens when the line between real and synthetic is practically non-existent? This study explores how a Qwen3-0.6B model, designed for mobile and edge deployment, can be fine-tuned to generate JSON-schematic data for workout and fitness. It shows that even smaller models are well-enough to design parsable output, ready to be used by an application - thus hiding its AI trademarks. In the future, studies on even smaller models may be appropriate, as well as discussions regarding its ethics.}},
author = {{Angenius, Simon}},
language = {{eng}},
note = {{Student Paper}},
title = {{Is Generative AI Hiding in Your Phone? Fine-tuning SLMs to Generate JSON-Adherent Output}},
year = {{2026}},
}