Quacer-C- Quantitative Certification of Knowledge Comprehension in LLMs
QuaCer-C is a framework for quantitatively certifying the knowledge comprehension capabilities of Large Language Models (LLMs). Using the structured nature of knowledge graphs, we are able to derive specifications for reasoning over unstructured data like text providing a way to formally understanding reasoning using exact confidence intervals.
Authors: Isha Chaudhary, Vedaant Jain, Gagandeep Singh
Sep 10, 2024
