Systematic Study of ZK-Friendly Quantization for Verifiable LLM Inference

This research evaluates how quantization methods impact the arithmetic complexity and proving costs of Zero-Knowledge (ZK) proofs for LLM inference. It identifies the relationship between quantization schemes and the underlying finite field operations required for verifiable privacy and governance.