Ordinality in Discrete-level Question Difficulty Estimation: Introducing Balanced DRPS and OrderedLogitNN

Arthur Thuy, Ekaterina Loginova, Dries F. Benoit · 2025

Read Ordinality in Discrete-level Question Difficulty Estimation: Introducing Balanced DRPS and OrderedLogitNN by Arthur Thuy, Ekaterina Loginova, Dries F. Benoit (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-2507-00736.txt.

Description

Arthur Thuy and co-authors argue that question difficulty should be treated as an ordered scale in this 2025 paper. Difficulty levels run from easiest to hardest, so order carries real meaning. Most past systems, though, treated the levels as plain classes or rough scores.

That habit also made study results hard to compare across papers. The team introduce a balanced scoring rule that respects level order. They also present a neural model built for ordered outcomes. Both tools aim to reward near misses over wild errors. Test makers can use them to sort questions more fairly. Better difficulty labels help exams match the students who take them.

Catalog record

CREATORArthur Thuy, Ekaterina Loginova, Dries F. Benoit
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-00736

Browse the full library · About the Tower of Babel