Records found: 3
A recent EPFL study benchmarks multimodal foundation models like GPT-4o on core vision tasks, revealing strengths in semantic understanding but highlighting gaps compared to specialized vision models.
MMaDA is a novel unified multimodal diffusion model that excels in textual reasoning, visual understanding, and image generation, outperforming existing systems across multiple benchmarks.
Skywork AI introduces R1V2, a cutting-edge multimodal reasoning model that blends hybrid reinforcement learning techniques to improve specialized reasoning and generalization, outperforming many open-source and proprietary models.