iheallab/Llama-Clin-REACT-70B
macontreras98/Llama-Clin-REACT-70B is a 70 billion parameter large language model, fine-tuned by macontreras98 based on Meta's Llama 3.3 70B Instruct. This model is specifically optimized for clinical reasoning tasks, particularly within the context of ICU decision-making, utilizing the ICU-REACT framework. It demonstrates superior performance on various clinical reasoning benchmarks compared to its base model and other open medical LLMs at a similar scale, making it suitable for research in medical AI.
Loading preview...
Llama-Clin-REACT-70B: Clinical Reasoning LLM
Llama-Clin-REACT-70B is a 70 billion parameter large language model developed by macontreras98, fine-tuned from the meta-llama/Llama-3.3-70B-Instruct base model. Its primary focus is on clinical reasoning, specifically designed to process and respond to clinically relevant information, particularly for Intensive Care Unit (ICU) decision-making using the ICU-REACT framework.
Key Capabilities and Performance
This model excels in various clinical reasoning benchmarks, outperforming its Llama 3.3 70B Instruct backbone and other 70B-scale open medical models. Key benchmark results (mean primary scores as percentages) include:
- ICU-REACT (n=71): 45.0%
- SCT-Bench (n=174): 67.5%
- ER-Reason (n=72): 50.9%
- MedRBench (n=1254): 46.0%
- VivaBench (n=934): 27.5%
It achieves a macro average of 47.4% across these benchmarks, indicating strong performance in diverse clinical reasoning scenarios. The model is released as a merged, full-parameter checkpoint, making it ready for inference.
Intended Use and Limitations
Clin-REACT is intended for research in clinical reasoning and medical AI. It is crucial to understand that this model is not a medical device and should not be used as a substitute for professional clinical judgment, diagnosis, or treatment decisions. Users are responsible for evaluating its suitability for their specific settings and ensuring compliance with all relevant privacy, security, and regulatory requirements.