Benchmarking Multi-Modal LLMs for Testing Visual Deep Learning Systems Through the Lens of Image Mutation

Visual deep learning (VDL) systems have shown significant success in real-world applications like image recognition, object detection, and autonomous driving. To evaluate the reliability of VDL, a mainstream approach is software testing, which requires diverse and controllable mutations over image semantics. The rapid development of multi-modal large language models (MLLMs) has introduced revolutionary image mutation potentials through instruction-driven methods. Users can now freely describe desired mutations and let MLLMs generate the mutated images. However, the quality of MLLM-produced test inputs in VDL testing remains largely unexplored. We present the first study, aiming to assess MLLMs' adequacy from 1) the semantic validity of MLLM mutated images, 2) the alignment of MLLM mutated images with their text instructions (prompts), 3) the faithfulness of how different mutations preserve semantics that are ought to remain unchanged, and 4) the effectiveness of detecting VDL faults. With large-scale human studies and quantitative evaluations, we identify MLLM's promising potentials in expanding the covered semantics of image mutations. Notably, while SoTA MLLMs (e.g., GPT-4V) fail to support or perform worse in editing existing semantics in images (as in traditional mutations like rotation), they generate high-quality test inputs using "semantic-additive" mutations (e.g., "dress a dog with clothes"), which bring extra semantics to images; these were infeasible for past approaches. Hence, we view MLLM-based mutations as a vital complement to traditional mutations, and advocate future VDL testing tasks to combine MLLM-based methods and traditional image mutations for comprehensive and reliable testing.