MIA
Emerging1papers using it
2025first seen
The 'MIA' dataset/benchmark is a challenging and diverse multi-modal instruction-following evaluation tool that assesses how well Multi-modal Large Language Models (MLLMs) understand and execute user instructions, incorporating both output response constraints and input image perception constraints.