SEP 9, 2026 · PREPRINT
Maverick: Private and Verifiable LLM Inference Made Practical via Matrix-Vector Multiplication Delegation
arXiv
An early-stage prototype implementation of a novel cryptographic protocol for LLM inference, demonstrating feasibility and throughput gains but lacking peer review, real-world deployment validation, or comparison against established privacy-preserving inference methods.
Reported
Throughput gain (1 client thread,…up to 17x
Throughput gain (1 client thread,…45x
Throughput gain (1 client thread,…44x