visual-task-semantics

Tag

Cards List
#visual-task-semantics

When Prompts Become Pixels: Prompt-Region Grounding for Multimodal Reasoning

arXiv cs.AI · 6d ago Cached

This paper introduces Visualized Task Semantics (VTS), a controlled intervention to study when multimodal LLMs are asked questions in the image rather than text. It reveals a consistent accuracy drop across all tested models and tasks, and proposes prompt-region grounding to recover the clean task representation, improving VTS accuracy from 58.0 to 66.3.

0 favorites 0 likes
← Back to home

Submit Feedback