Visual Program Distillation: Distilling Tools and Programmatic Reasoning into Vision-Language Models

Yushi Hu, Otilia Stretcu, Chun-Ta Lu, Krishnamurthy Viswanathan, Kenji Hata, Enming Luo, Ranjay Krishna, Ariel Fuxman; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 9590-9601

Solving complex visual tasks such as "Who invented the musical instrument on the right?" involves a composition of skills: understanding space recognizing instruments and also retrieving prior knowledge. Recent work shows promise by decomposing such tasks using a large language model (LLM) into an executable program that invokes specialized vision models. However generated programs are error-prone: they omit necessary steps include spurious ones and are unable to recover when the specialized models give incorrect outputs. Moreover they require loading multiple models incurring high latency and computation costs. We propose Visual Program Distillation (VPD) an instruction-tuning framework that produces a vision-language model (VLM) capable of solving complex visual tasks with a single forward pass. VPD distills the reasoning ability of LLMs by using them to sample multiple candidate programs which are then executed and verified to identify the correct one. It translates each correct program into a language description of the reasoning steps which are then distilled into a VLM. Extensive experiments show that VPD improves the VLM's ability to count understand spatial relations and reason compositionally. Our VPD-trained PaLI-X outperforms all prior VLMs achieving state-of-the-art performance across complex vision tasks including MMBench OK-VQA A-OKVQA TallyQA POPE and Hateful Memes. An evaluation with human annotators also confirms that VPD improves model response factuality and consistency. Finally experiments on content moderation demonstrate that VPD is also helpful for adaptation to real-world applications with limited data.