ChronoSense: Exploring Temporal Understanding in Large Language Models with Time Intervals of Events
Duygu Sezen Islakoglu *, Jan-Christoph Kalo * · 2025
Read ChronoSense: Exploring Temporal Understanding in Large Language Models with Time Intervals of Events by Duygu Sezen Islakoglu *, Jan-Christoph Kalo * (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-2501-03040.txt.
Description
Duygu Sezen Islakoglu and Jan-Christoph Kalo present ChronoSense, a benchmark for how language models grasp time spans. Their paper from twenty twenty-five notes that models still stumble on reasoning and arithmetic. Time reasoning is a core part of language sense, yet tests of interval links stay rare. Allen's framework names those links: before, after, during, and more.
The set holds sixteen tasks. Some ask which Allen link joins two events. Others ask for time arithmetic. Events are both abstract and drawn from Wikidata's real records. Seven recent models were tested. Scores swung widely, even between mirror-image links. The team suspects rote recall props up some right answers. Low marks overall mark time sense as a weak spot. Data and code are public for future work.
Catalog record
| CREATOR | Duygu Sezen Islakoglu *, Jan-Christoph Kalo * |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-03040