IBI-CAAI/MELT-TinyLlama-1.1B-Chat-v1.0

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1.1BQuant:BF16Ctx Length:2kLicense:apache-2.0Architecture:Transformer0.0K Open Weights Warm

The IBI-CAAI/MELT-TinyLlama-1.1B-Chat-v1.0 is a 1.1 billion parameter generative text model developed by the Center for Applied AI and funded by the Institute of Biomedical Informatics. It is fine-tuned from TinyLlama-1.1B-Chat-v1.0 using extensive medical domain data, demonstrating a 13.76% average improvement over its base model on medical benchmarks like USMLE, Indian AIIMS, and NEET. This model is specifically optimized for medical question-answering and chat formats, intended for research purposes in the medical field.

Loading preview...

Overview

MELT-TinyLlama-1.1B-Chat-v1.0 is a 1.1 billion parameter Large Language Model (LLM) developed by the Center for Applied AI and funded by the Institute of Biomedical Informatics. It is a specialized version of TinyLlama-1.1B-Chat-v1.0, fine-tuned extensively on a wide range of publicly available medical text, chat, Q&A, and instruction data. The model is designed to excel in the medical domain, demonstrating significant performance improvements over its base model.

Key Capabilities & Performance

  • Medical Domain Specialization: Trained on diverse medical datasets including Expert Med, MedQA, MedMCQA, LiveQA, and various medical flashcards and conversational data.
  • Enhanced Medical Q&A: Achieves an average 13.76% improvement over TinyLlama-1.1B-Chat-v1.0 across three key medical benchmarks: USMLE, Indian AIIMS, and NEET medical examinations.
  • Chat and QA Format Optimized: Best suited for prompts structured in question-answering or chat formats, making it practical for interactive medical information retrieval.

Intended Use Cases

  • Medical Research: Primarily intended for research purposes within the medical domain.
  • Educational Support: Can be used to explore medical concepts and answer questions based on its specialized training.

Limitations and Important Considerations

  • Research Use Only: This model is strictly for research and should not be used for providing medical advice or clinical decision-making.
  • Potential Biases: Trained on publicly available data which may contain biases or inaccuracies; the training and evaluation datasets have not been fully vetted for content or accuracy.
  • Disclaimer: Users must exercise discretion as the model may produce inaccurate or outdated information. The developers disclaim liability for any consequences arising from its use.