Tempora: Characterising the Time-Contingent Utility of Online Test-Time Adaptation
Sudarshan Sreeram, Young D. Kwon, Cecilia Mascolo · 2026
Read Tempora: Characterising the Time-Contingent Utility of Online Test-Time Adaptation by Sudarshan Sreeram, Young D. Kwon, Cecilia Mascolo (2026) 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-2602-06136.txt.
Description
Models that adapt on the fly to new data can fix their own drift, but adaptation takes time, and late answers may be useless. Sudarshan Sreeram and 2 co-authors present Tempora, a framework that grades adaptation methods under real time pressure. Standard leaderboards assume unlimited processing time.
Tempora scores methods with three utility measures: hard deadlines for streams, decaying value for interactive use, and budgets for cost-limited settings. Across hundreds of runs with many methods, data sets, and machines, rankings kept changing. The best method under no time pressure often lost once delays counted. No single winner emerged, so the framework helps users pick based on their own time limits.
Catalog record
| CREATOR | Sudarshan Sreeram, Young D. Kwon, Cecilia Mascolo |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 68 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-06136