Jul 2026
Towards Grounded GI Endoscopy VQA via Multi-Task Learning on Small VLMs
A simple multi-task fine-tuning recipe that constructs auxiliary grounding and description tasks from an existing VQA dataset with minimal additional annotation is proposed, showing consistent accuracy gains together with improved implicit alignment between answer tokens and the relevant image region.
Itbaan Safwan, Ramail Khan, Muhammad Annas Shaikh et al.
· arXiv.org · 0 citations