m3rg-iitd/llamat-3-chat
LLaMat-3-Chat is an 8 billion parameter large language model developed by M3RG, IIT Delhi & DAIR, IIT Delhi, specifically fine-tuned from LLaMat-3 for materials research. It excels as an AI copilot for tasks like information extraction from material science text and tabular data, scientific data processing, and analyzing experimental results. The model is optimized for domain-specific instruction following and parsing data for research tasks within the material science field, building upon a base model that underwent continued pretraining on material science data.
Loading preview...
LLaMat-3-Chat: An AI Copilot for Materials Research
LLaMat-3-Chat is an 8 billion parameter large language model (LLM) developed by M3RG, IIT Delhi & DAIR, IIT Delhi. It is a specialized version of LLaMat-3, which itself is a continued pretraining of LLaMA-3 on extensive material science data. This model is specifically fine-tuned to act as an AI copilot for materials research, focusing on scientific data processing and interpretation.
Key Capabilities
- Domain-Specific Expertise: Optimized for understanding and processing instructions within the material science domain, leveraging its pretraining on material science tokens.
- Information Extraction: Proficient in extracting structured information from material science texts and tabular data.
- Table Understanding: Designed to interpret and parse data from scientific tables.
- Instruction Following: Enhanced for precise instruction following in research-oriented tasks.
Intended Use Cases
LLaMat-3-Chat is designed to assist researchers, scientists, and industry professionals in:
- Analyzing experimental results and processing large datasets.
- Supporting literature reviews and knowledge discovery in material science.
- Handling research-driven natural language queries related to materials.
This model was trained on a curated corpus including material science research papers, community discourse, and specific datasets like Redpajama, Openorca, mathQA, and MatSciNLP. Further details on its performance and comparison with other models are available in the associated paper: Foundational Large Language Models for Materials Research.