Papers2

#vision-language alignment

LatentLens: Revealing Highly Interpretable Visual Tokens in LLMs

Benno Krojer, Shravan Nayak et al.Jan 31arXiv

LatentLens is a simple, training-free way to translate what a model "sees" in image patches into clear words and phrases.

#LatentLens#visual tokens#contextual embeddings

Robust-R1: Degradation-Aware Reasoning for Robust Visual Understanding

Beginner

Jiaqi Tang, Jianmin Chen et al.Dec 19arXiv

Robust-R1 teaches vision-language models to notice how a picture is damaged, think through what that damage hides, and then answer as if the picture were clear.

#Robust-R1#degradation-aware reasoning#multimodal large language models