[HUGGINGFACE]score: 0.36
MM-CreativityBench Tests Affordance-Grounded Creative Tool Use in Multimodal Models
May 24, 2026
MM-CreativityBench evaluates whether LMMs can identify non-obvious repurposing of scene elements under physical constraints, going beyond pattern recognition to affordance-grounded creative problem solving. Each instance pairs a scenario image with structured views of candidate objects requiring physically feasible solutions.
paper
HOW THIS AFFECTS YOU
●
researcherThis benchmark isolates a capability gap in current LMMs that standard VQA and reasoning benchmarks do not surface, useful for evaluating embodied and creative reasoning.