SoC: Semantic Orthogonal Calibration for Test-Time Prompt Tuning
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911723935498240 |
|---|---|
| author | Fillioux, Leo Chakraborty, Omprakash Ayed, Ismail Ben Cournède, Paul-Henry Christodoulidis, Stergios Vakalopoulou, Maria Dolz, Jose |
| author_facet | Fillioux, Leo Chakraborty, Omprakash Ayed, Ismail Ben Cournède, Paul-Henry Christodoulidis, Stergios Vakalopoulou, Maria Dolz, Jose |
| contents | With the increasing adoption of vision-language models (VLMs) in critical decision-making systems such as healthcare or autonomous driving, the calibration of their uncertainty estimates becomes paramount. Yet, this dimension has been largely underexplored in the VLM test-time prompt-tuning (TPT) literature, which has predominantly focused on improving their discriminative performance. Recent state-of-the-art advocates for enforcing full orthogonality over pairs of text prompt embeddings to enhance separability, and therefore calibration. Nevertheless, as we theoretically show in this work, the inherent gradients from fully orthogonal constraints will strongly push semantically related classes away, ultimately making the model overconfident. Based on our findings, we propose Semantic Orthogonal Calibration (SoC), a Huber-based regularizer that enforces smooth prototype separation while preserving semantic proximity, thereby improving calibration compared to prior orthogonality-based approaches. Across a comprehensive empirical validation, we demonstrate that SoC consistently improves calibration performance, while also maintaining competitive discriminative capabilities. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_08617 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SoC: Semantic Orthogonal Calibration for Test-Time Prompt Tuning Fillioux, Leo Chakraborty, Omprakash Ayed, Ismail Ben Cournède, Paul-Henry Christodoulidis, Stergios Vakalopoulou, Maria Dolz, Jose Computer Vision and Pattern Recognition With the increasing adoption of vision-language models (VLMs) in critical decision-making systems such as healthcare or autonomous driving, the calibration of their uncertainty estimates becomes paramount. Yet, this dimension has been largely underexplored in the VLM test-time prompt-tuning (TPT) literature, which has predominantly focused on improving their discriminative performance. Recent state-of-the-art advocates for enforcing full orthogonality over pairs of text prompt embeddings to enhance separability, and therefore calibration. Nevertheless, as we theoretically show in this work, the inherent gradients from fully orthogonal constraints will strongly push semantically related classes away, ultimately making the model overconfident. Based on our findings, we propose Semantic Orthogonal Calibration (SoC), a Huber-based regularizer that enforces smooth prototype separation while preserving semantic proximity, thereby improving calibration compared to prior orthogonality-based approaches. Across a comprehensive empirical validation, we demonstrate that SoC consistently improves calibration performance, while also maintaining competitive discriminative capabilities. |
| title | SoC: Semantic Orthogonal Calibration for Test-Time Prompt Tuning |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.08617 |