Advancing Creative Physical Intelligence in Large Multimodal Models

2026-05-25

AI Summary

Existing large models struggle to generate non-obvious yet physically feasible tool-use strategies in open-world settings due to insufficient grounding in visual and physical constraints. To address this, this work introduces MM-CreativityBench, the first benchmark for systematically evaluating embodied creativity in large models. The proposed approach incorporates a functional attribute alignment mechanism that leverages a functional knowledge base for supervision, multi-view structured scene representations, multi-turn interactive reasoning, and preference learning via Direct Preference Optimization. This framework significantly improves the model’s accuracy in selecting appropriate objects and parts, effectively mitigates hallucination and grounding errors, and consistently enhances performance across creative physical reasoning tasks.

Abstract

Large multimodal models (LMMs) have rapidly advanced in perception and reasoning; however, it remains unclear whether these capabilities generalize to discovering visually grounded solutions in open-ended environments, beyond pattern recognition. In such settings, intelligence requires more than answering well-posed questions: it involves identifying how elements in a scene can be repurposed in non-obvious yet physically feasible ways. This form of creative problem-solving is central to human intelligence, but remains largely untested in current benchmarks. To evaluate this ability, we introduce MM-CreativityBench, a benchmark for affordance-grounded creative tool use in visually rich, physically constrained environments. Each instance presents a scenario image with structured views of candidate entities and their parts, enabling fine-grained, interactive evaluation of how models iteratively inspect the scene, identify relevant affordances, and compose visually and physically grounded solutions. Our experiments show that current LMMs often fall short, not due to lack of generative capability, but because they do not sustain grounded exploration. Models often overlook relevant entities, under-examine critical parts, or hallucinate attributes not grounded in the image. Motivated by this failure mode, we propose affordance-grounded alignment, which casts creative tool use as a preference learning problem. Using Direct Preference Optimization, we encourage models to prefer attribute-affordance reasoning grounded in visual evidence over hallucinated alternatives. In addition, we incorporate supervision derived from an affordance knowledge base to guide broader entity exploration and multi-turn planning. Our results show consistent gains in selecting the correct entities and parts, while substantially reducing hallucination and grounding-related errors.

Problem

• creative problem-solving
• affordance grounding
• large multimodal models
• physical intelligence
• visual reasoning

Innovation

• affordance-grounded alignment
• creative tool use
• multimodal reasoning
• preference learning
• visual grounding