RoboCrafter-QA demonstrates how Large Language Models can effectively evaluate and select optimal soft robot designs by applying embodied reasoning principles.

• LLMs can analyze complex design trade-offs involving materials, morphology, and mechanics
• The benchmark tests LLMs' ability to understand physical constraints and engineering principles
• Results show LLMs can serve as effective design partners in robotics development
• This approach reduces the need for costly trial-and-error experimentation

This research creates a path toward AI-assisted engineering workflows where LLMs help experts navigate complex design spaces in robotics, potentially accelerating innovation while reducing development costs.

Large Language Models as Natural Selector for Embodied Soft Robot Design