Welcome

Welcome to the kibad-llm docs.

This is an information extraction project focused on supporting the extension of the Literaturdatenbank Faktencheck Artenvielfalt. It uses LLMs to extract structured information from scientific literature PDFs.

Using the project

• Quickstart — Cheat sheet: setup, first prediction and evaluation, architecture overview, and repo conventions
• Usage instructions — Step-by-step guide: PDF download, database conversion, prediction, and evaluation
• LLM usage instructions — How to host and run language models on the cluster
• Faktencheck database instructions — Setting up the PostgreSQL database via Podman
• Evaluation dashboard — Interactive browser tool for visualizing and comparing prediction results

Contributing

• Overview — PR workflow, branch conventions, and CI requirements
• Source code guidelines — Coding standards, testing, and docstring conventions
• Experiment guidelines — How to plan, run, and document experiments

Code reference

Auto-generated documentation from Python docstrings is available in the KIBA-D code reference.