Yingying Yan

Image Quality Assessment (IQA) in open-world environments remains challenging due to limited generalization and interpretability. Recent approaches based on multimodal large language models (MLLMs) introduce textual reasoning for quality prediction, yet their judgments rely heavily on semantically biased internal representations, making them insensitive to low-level perceptual degradations.

1 Introduction The pursuit of interfacial stable lithium metal batteries (LMBs) has been a longstanding challenge in the field, particularly since metallic lithium replaces graphite as an anode material.