SoC: Semantic Orthogonal Calibration for Test-Time Prompt Tuning
Leo Fillioux, Omprakash Chakraborty, Ismail Ben Ayed, Paul-Henry Cournède, Stergios Christodoulidis, Maria Vakalopoulou, Jose Dolz · 2026
Read SoC: Semantic Orthogonal Calibration for Test-Time Prompt Tuning by Leo Fillioux, Omprakash Chakraborty, Ismail Ben Ayed, Paul-Henry Cournède, Stergios Christodoulidis, Maria Vakalopoulou, Jose Dolz (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-2601-08617.txt.
Description
Leo Fillioux and four co-authors wrote a 2026 arXiv paper on how sure a vision-language model should be of its own replies. Such models now serve in health care and self-driving cars, where a wrong reply can harm. Test-time prompt tuning, or TPT, adapts a model at the time of use by tuning its prompt. Past TPT work has chased sharp picks, not trust in them. The authors set out Semantic Orthogonal Calibration, or SoC. It aims at the trust side that past work left aside.
Catalog record
| CREATOR | Leo Fillioux, Omprakash Chakraborty, Ismail Ben Ayed, Paul-Henry Cournède, Stergios Christodoulidis, Maria Vakalopoulou, Jose Dolz |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-08617