Large Language Models for Real-World Nutrition Assessment: Structured Prompts, Multi-Model Validation and Expert Oversight
Aia Ase, Jacek Borowicz, Kamil Rakocy, Barbara Piekarska · 2025
Read Large Language Models for Real-World Nutrition Assessment: Structured Prompts, Multi-Model Validation and Expert Oversight by Aia Ase, Jacek Borowicz, Kamil Rakocy, Barbara Piekarska (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12788131.txt.
Description
Checking what people eat is slow and error-prone, and Aia Ase and co-authors test whether large language models can do the job in Polish, a language with complex word forms. They ran nearly two thousand food items from a long-term care home through three leading models. Each item was classified under health schemes with either a strict two-step prompt or a simpler one-step prompt, and results were compared with two human experts.
All three models agreed with the experts ninety to ninety-four percent of the time. The strict prompt caught nearly every unhealthy item but flagged too many healthy ones, while the simple prompt was more balanced. The authors conclude these models reach near-expert accuracy and can shift human effort from first-pass sorting to checking. They stress that expert oversight stays essential and testing in each language matters.
Catalog record
| CREATOR | Aia Ase, Jacek Borowicz, Kamil Rakocy, Barbara Piekarska |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |