Inverse-LLaVA: Eliminating Alignment Pre-training Through Text-to-Vision Mapping Paper • 2508.12466 • Published Aug 17, 2025 • 9
