CrystalReasoner/Qwen2.5-3B-CrysReas-PLaIDWyckoff
CrystalReasoner/Qwen2.5-3B-CrysReas-PLaIDWyckoff is a 3.1 billion parameter Qwen2.5-based language model developed by CrystalReasoner. This model is an end-to-end framework for generating crystal structures from natural language instructions. It utilizes supervised fine-tuning, reasoning traces for crystallographic priors, and reinforcement learning with verifiable rewards to enhance validity, stability, and property conditioning. Its primary strength lies in property-conditioned crystal structure generation, making it suitable for materials science and chemistry applications.
Loading preview...
CrystalReasoner/Qwen2.5-3B-CrysReas-PLaIDWyckoff Overview
This model, developed by CrystalReasoner, is a 3.1 billion parameter variant of the Qwen2.5 architecture, specifically designed for crystal structure generation. It operates as an end-to-end framework, translating natural language instructions into detailed crystal structures.
Key Capabilities
- Crystal Structure Generation: Generates crystal structures directly from natural language descriptions.
- Property-Conditioned Generation: Capable of generating structures based on specified material properties, such as chemical formula and bulk modulus.
- Reasoning Integration: Incorporates "thinking traces" to embed crystallographic and physical priors before generating atomic coordinates.
- Reinforcement Learning (RL): Utilizes RL with verifiable rewards to improve the validity, stability, and property conditioning of generated structures.
- Pymatgen Compatibility: Generated structures can be easily converted into
pymatgen.core.Structureobjects for further analysis.
Good For
- Researchers and engineers in materials science, chemistry, and crystallography.
- Automated design and discovery of novel materials with specific properties.
- Generating detailed crystal lattice parameters and atomic coordinates from high-level descriptions.