New GRAFT Method Uses Graph-Regularized Prototype Fields for Test-Time Adaptation of Vision-Language Models
A new arXiv paper introduces GRAFT, a method using graph-regularized prototype fields to adapt vision-language model zero-shot classifiers with unlabeled test data, addressing gaps in existing transductive and test-time adaptation approaches.
Read the full story at arXiv cs.CV ↗
Timeline · 1 report
- 2026-11-19 05:00 · arXiv cs.CV
GRAFT: Graph-Regularized Prototype Fields for Realistic Test-Time Adaptation of VLMs