lucidrains/clinical-calculator-tooluse

Explorations into training LLMs to use clinical calculators from patient history, using open sourced models. Will start with Wells' Criteria

What it solves

This project explores how to train Large Language Models (LLMs) to accurately use clinical calculators—specialized medical tools used to predict risk or diagnose conditions—by extracting necessary information from a patient's history.

How it works

It utilizes open-source models to experiment with tool-use capabilities, specifically focusing on training the model to identify the correct data points from medical records and input them into clinical calculators, starting with the Wells' Criteria for pulmonary embolism risk assessment.

Who it’s for

Researchers and developers interested in the intersection of medical AI, tool-use training for LLMs, and the application of open-source models in clinical decision support.

Highlights

  • Focuses on clinical calculator tool-use.
  • Uses open-source LLMs.
  • Initial implementation focuses on Wells' Criteria.

Related

  • Project
    simonw/llm

    A CLI tool and Python library for interacting with multiple LLMs, both via remote APIs and local installations, featuring built-in logging and structured data extraction.

  • Project
    OpenTSLM/OpenTSLM

    OpenTSLM is an open‑source library that adds native time‑series handling to Llama 3.2 and Gemma LLMs, enabling natural‑language prompting and reasoning over medical signals (ECG, EEG, accelerometer, etc.). It provides pre‑trained checkpoints, a simple `OpenTSLM` Python API, demo scripts for inference on several benchmark tasks, and a curriculum‑learning training pipeline covering MCQ, captioning, and chain‑of‑thought reasoning stages. Install via `pip install opentslm`, load models from Hugging‑Face, and fine‑tune or evaluate using the supplied scripts. Licensed under MIT.

  • Dispatch
    The Open Medical-LLM Leaderboard: Benchmarking Large Language Models in Healthcare

    Hugging Face has introduced the Open Medical-LLM Leaderboard, a standardized platform to evaluate and compare the performance of LLMs across diverse medical datasets to improve reliability and patient safety.

  • Project
    pengzhangzhi/Open-dLLM

    A complete open-source framework for diffusion-based large language models, providing the entire pipeline from pretraining and datasets to evaluation and inference.

  • Project
    lyogavin/airllm

    AirLLM is a Python library that streams model layers from disk so only one layer lives on the GPU at a time, enabling inference of 70 B‑plus LLMs on 4‑12 GB GPUs and fine‑tuning of 125 B models on < 6 GB. It supports most open‑source LLM families, optional 4‑/8‑bit block‑wise compression for speed, and works on Linux, Windows, and Apple Silicon.