Efficient Scientific Full Text Classification: The Case of EICAT Impact Assessments

Marc Brinner, Sina Zarrieß · 2025

Read Efficient Scientific Full Text Classification: The Case of EICAT Impact Assessments by Marc Brinner, Sina Zarrieß (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/arxiv-2502-06551.txt.

Description

Marc Brinner and Sina Zarrieß sort full science papers with lean models in a 2025 study. Their case is EICAT, a system that rates the harm done by invasive species. Human raters read whole papers to back each rating, a slow craft. The team asks how small models can sort these long texts well. They also test a local large language model, Llama-3.1 8B.

They built a data set of invasion biology papers matched to public EICAT ratings. Their trick is to feed the sorter only picked sentences, not the whole text. Those picks can come from human notes, model-made notes, or scores that show why a model chose a class. Shorter inputs beat full text even against models built for long input. Repeated sampling of short inputs helped more.

Catalog record

CREATORMarc Brinner, Sina Zarrieß
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-06551

Browse the full library · About the Tower of Babel