Abstract:VLMs have become increasingly proficient at a range of computer vision tasks, such as visual question answering and object detection. This includes increasingly strong capabilities in the domain of art, from analyzing artwork to generation of art. In an interdisciplinary collaboration between computer scientists and art historians, we characterize the mechanisms underlying VLMs' ability to predict artistic style and assess the extent to which they align with the criteria art historians use to reason about artistic style. We employ a latent-space decomposition approach to identify concepts that drive art style prediction and conduct quantitative evaluations, causal analysis and assessment by art historians. Our findings indicate that 73% of the extracted concepts are judged by art historians to exhibit a coherent and semantically meaningful visual feature and 90% of concepts used to predict style of a given artwork were judged relevant. In cases where an irrelevant concept was used to successfully predict style, art historians identified possible reasons for its success; for example, the model might "understand" a concept in more formal terms, such as dark/light contrasts.




Abstract:We propose a two-step approach for detecting differences in the style of images across sources of differing cultural affinity, where images are first clustered into finer visual themes based on content before their aesthetic features are compared. We test this approach on 2,400 YouTube video thumbnails taken equally from two U.S. and two Chinese YouTube channels, and relating equally to COVID-19 and the Ukraine conflict. Our results suggest that while Chinese thumbnails are less formal and more candid, U.S. channels tend to use more deliberate, proper photographs as thumbnails. In particular, U.S. thumbnails are less colorful, more saturated, darker, more finely detailed, less symmetric, sparser, less varied, and more up close and personal than Chinese thumbnails. We suggest that most of these differences reflect cultural preferences, and that our methods and observations can serve as a baseline against which suspected visual propaganda can be computed and compared.
Abstract:LLMs can be unpredictable, as even slight alterations to the prompt can cause the output to change in unexpected ways. Thus, the ability of models to accurately explain their behavior is critical, especially in high-stakes settings. One approach for evaluating explanations is counterfactual simulatability, how well an explanation allows users to infer the model's output on related counterfactuals. Counterfactual simulatability has been previously studied for yes/no question answering tasks. We provide a general framework for extending this method to generation tasks, using news summarization and medical suggestion as example use cases. We find that while LLM explanations do enable users to better predict LLM outputs on counterfactuals in the summarization setting, there is significant room for improvement for medical suggestion. Furthermore, our results suggest that the evaluation for counterfactual simulatability may be more appropriate for skill-based tasks as opposed to knowledge-based tasks.