Evaluating and Enhancing Negation Comprehension in Remote Sensing MLLMs
Researchers have introduced RS-Neg, a new benchmark designed to evaluate and improve the negation comprehension abilities of Multimodal Large Language Models (MLLMs) in remote sensing tasks. Current advanced MLLMs exhibit significant limitations in understanding negation, leading to hallucinations and performance degradation. To address this, a novel test-time learning method called NeFo has been proposed, which incorporates the logical role of negation into model optimization and shows strong generalization capabilities with minimal unlabeled data. AI
IMPACT Enhances AI's ability to accurately interpret negative statements in critical applications like emergency response.