Defensive evaluation of refusal calibration when LLMs interpret science foundation model outputs, with reproducible harnesses and 24.3K outcome records.
-
Updated
Jul 12, 2026 - Python
Defensive evaluation of refusal calibration when LLMs interpret science foundation model outputs, with reproducible harnesses and 24.3K outcome records.
Production-grade LLM evaluation framework measuring model behavior across 5 dimensions with human-vs-LLM judge agreement validation and Cohen's Kappa scoring
Policy and evaluation framework for ChemBio risk classification in advanced AI systems.
Add a description, image, and links to the refusal-calibration topic page so that developers can more easily learn about it.
To associate your repository with the refusal-calibration topic, visit your repo's landing page and select "manage topics."