Loading
Loading articles
By Ferhat Erata

AI summary
This article introduces a semantic diffing approach for formal policy verification using anti-unification over e-graphs. By accounting for algebraic laws and domain-specific identities, this technique isolates meaningful behavioral differences in code or policy expressions while ignoring superficial syntactic noise. The approach is particularly useful for validating LLM-generated artifacts by identifying where multiple candidate formalizations disagree.
Key points