A Surprising Failure? Multimodal LLMs and the NLVR Challenge
This highlights a critical weakness in current MLLMs for tasks requiring precise compositional reasoning, which is important for applications like robotics and AI assistants.
The study evaluated three state-of-the-art multimodal LLMs (GPT-4V, Gemini Pro, IDEFICS) on the NLVR task, which tests compositional and spatial reasoning, and found they performed poorly despite their strong overall capabilities.
This study evaluates three state-of-the-art MLLMs -- GPT-4V, Gemini Pro, and the open-source model IDEFICS -- on the compositional natural language vision reasoning task NLVR. Given a human-written sentence paired with a synthetic image, this task requires the model to determine the truth value of the sentence with respect to the image. Despite the strong performance demonstrated by these models, we observe they perform poorly on NLVR, which was constructed to require compositional and spatial reasoning, and to be robust for semantic and systematic biases.