This research demonstrates how Large Language Models can effectively translate between molecular structures and natural language descriptions, eliminating the need for domain-specific pre-training.

• Achieves state-of-the-art performance on molecule-caption translation tasks
• Uses in-context learning with carefully designed prompts
• Works with mainstream LLMs without requiring specialized molecular training
• Maintains reasonable computational efficiency compared to previous approaches

This breakthrough accelerates medical research and drug discovery by making molecular information more accessible and interpretable to researchers without requiring specialized AI systems.

Large Language Models are In-Context Molecule Learners