#depth prediction

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.