Jul 2026
RadSight: Towards Perceptually Reliable Multimodal Radiology Image Understanding
RadSight is proposed, a perception-driven MLLM built upon a dual 2D/3D encoder architecture that preserves native imaging spatial structures that achieves consistent improvements on public 2D and 3D medical benchmarks, further demonstrating that robust low-level visual perception is a critical foundation for reliable clinical understanding.
Jianqi Liu, Weiwei Cao, Wanxing Chang et al.
· arXiv.org · 0 citations