Search papers, labs, and topics across Lattice.
The paper introduces CHAIN, a new interactive benchmark for evaluating vision reasoning in dynamic, physics-driven 3D environments, focusing on tasks like interlocking puzzles and 3D packing. CHAIN assesses a model's ability to understand, plan, and execute structured action sequences grounded in physical constraints, moving beyond static perception-based evaluations. Experiments with state-of-the-art VLMs and diffusion models reveal limitations in internalizing physical structure, causal constraints, and translating perception into effective actions for long-horizon planning.
Current VLMs can ace image quizzes, but completely fumble when asked to stack blocks in a physically plausible way, revealing a critical gap in understanding real-world physics.
Understanding the physical structure is essential for real-world applications such as embodied agents, interactive design, and long-horizon manipulation. Yet, prevailing Vision-Language Model (VLM) evaluations still center on structure-agnostic, single-turn setups (e.g., VQA), which fail to assess agents'ability to reason about how geometry, contact, and support relations jointly constrain what actions are possible in a dynamic environment. To address this gap, we introduce the Causal Hierarchy of Actions and Interactions (CHAIN) benchmark, an interactive 3D, physics-driven testbed designed to evaluate whether models can understand, plan, and execute structured action sequences grounded in physical constraints. CHAIN shifts evaluation from passive perception to active problem solving, spanning tasks such as interlocking mechanical puzzles and 3D stacking and packing. We conduct a comprehensive study of state-of-the-art VLMs and diffusion-based models under unified interactive settings. Our results show that top-performing models still struggle to internalize physical structure and causal constraints, often failing to produce reliable long-horizon plans and cannot robustly translate perceived structure into effective actions. The project is available at https://social-ai-studio.github.io/CHAIN/.
