Tutorial
Semantic diffing of formal policies with anti-unification
This AWS Machine Learning Blog post introduces an open-source engine that computes semantic diffs of formal policies using anti-unification over e-graphs. It ignores syntactic changes covered by declared algebraic laws and domain rewrites, and localizes genuine disagreements as Variants nodes. The post demonstrates the approach on authorization policy examples and LLM autoformalization, showing how it reduces a textual diff of 18 flagged lines to two localized semantic bugs.
formal methodsLLM autoformalizationauthorization policies