BlenderGym: Benchmarking Foundational Model Systems for Graphics Editing
This work addresses the bottleneck in developing and evaluating VLMs for 3D graphics editing, which is crucial for applications like movie production and game design, by providing a new benchmark.
The authors tackled the lack of a comprehensive benchmark for vision-language models (VLMs) in 3D graphics editing by introducing BlenderGym, which evaluates VLM systems through code-based 3D reconstruction tasks and found that even state-of-the-art VLMs struggle with tasks easy for humans, while also showing that inference scaling techniques can optimize performance by strategically distributing compute between generation and verification.
3D graphics editing is crucial in applications like movie production and game design, yet it remains a time-consuming process that demands highly specialized domain expertise. Automating this process is challenging because graphical editing requires performing a variety of tasks, each requiring distinct skill sets. Recently, vision-language models (VLMs) have emerged as a powerful framework for automating the editing process, but their development and evaluation are bottlenecked by the lack of a comprehensive benchmark that requires human-level perception and presents real-world editing complexity. In this work, we present BlenderGym, the first comprehensive VLM system benchmark for 3D graphics editing. BlenderGym evaluates VLM systems through code-based 3D reconstruction tasks. We evaluate closed- and open-source VLM systems and observe that even the state-of-the-art VLM system struggles with tasks relatively easy for human Blender users. Enabled by BlenderGym, we study how inference scaling techniques impact VLM's performance on graphics editing tasks. Notably, our findings reveal that the verifier used to guide the scaling of generation can itself be improved through inference scaling, complementing recent insights on inference scaling of LLM generation in coding and math tasks. We further show that inference compute is not uniformly effective and can be optimized by strategically distributing it between generation and verification.