M-PFT: Angular Margin for Calibrated Prompt-Tuning of Fine-Tuned Vision-Language Models

Prompt-tuning has emerged as a parameter-efficient technique for adapting large-scale vision-language models (VLMs) to downstream tasks while keeping the pre-trained weights frozen.

By Shihab Aaqil Ahamed, Mohamed Ansaff, Nang Khin Myat Lay Phyu et al · 1-minute read