DUTIR-BioNLP/RexDrug-base

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Mar 8, 2026License:llama3.1Architecture:Transformer0.0K Cold

DUTIR-BioNLP/RexDrug-base is an 8 billion parameter supervised fine-tuned (SFT) base model, built upon the Llama-3.1-8B-Instruct architecture. Developed by DUTIR-BioNLP, it is specifically optimized for chain-of-thought reasoning in biomedical drug combination relation extraction. This model identifies beneficial, harmful, neutral/mixed, or no combination relationships between drugs from scientific literature, serving as the foundation for the RexDrug system.

Loading preview...

RexDrug-base: Specialized for Biomedical Drug Combination Extraction

DUTIR-BioNLP/RexDrug-base is an 8 billion parameter language model, fine-tuned from the Llama-3.1-8B-Instruct architecture. Its primary purpose is to serve as the supervised fine-tuning (SFT) base model for RexDrug, a system designed for chain-of-thought reasoning in biomedical drug combination relation extraction.

Key Capabilities

  • Biomedical Relation Extraction: Specializes in identifying relationships between drug combinations within biomedical literature.
  • Specific Relation Types: Capable of classifying relationships into four distinct categories:
    • POS: Beneficial interactions.
    • NEG: Harmful interactions.
    • COMB: Neutral or mixed interactions.
    • NO_COMB: Indicates no combination relationship.
  • Foundation for RexDrug: This model is intended to be used in conjunction with the RexDrug-adapter, a LoRA adapter trained via GRPO, to achieve its full functionality.

Good For

  • Researchers and developers working on automated analysis of biomedical texts.
  • Applications requiring precise extraction of drug interaction information from scientific papers.
  • Building specialized systems for pharmacovigilance or drug discovery support.