We present the first study on how Multimodal LLMs' (MLLMs) reasoning ability shall be elicited to evaluate the aesthetics of artworks. To facilitate this investigation, we construct MM-StyleBench, a novel high-quality dataset for benchmarking artistic stylization. We then develop a principled method for human preference modeling and perform a systematic correlation analysis between MLLMs' responses and human preference. Our experiments reveal an inherent hallucination issue of MLLMs in art evaluation, associated with response subjectivity. ArtCoT is proposed, demonstrating that art-specific task decomposition and the use of concrete language boost MLLMs' reasoning ability for aesthetics. Our findings offer valuable insights into MLLMs for art and can benefit a wide range of downstream applications, such as style transfer and artistic image generation. Code available at https://github.com/songrise/MLLM4Art.

NEW YORK: Two paintings up for auction in New York highlight a growing interest in artificial intelligence-created works - a technique that could transform how art is made and viewed but is also stirring up passionate debate. The art world was stunned last year when an AI painting sold for $432,500, and auctioneers are keen to further test demand for computer-generated works. "Art is a true reflection of what our society, what our environment responds to," said Max Moore of Sotheby's. "And so it's just a natural continuation of the progression of art," he added. Sotheby's will put two paintings by the French art collective Obvious up for sale on Thursday, including'Le Baron De Belamy'.