Gryphes Hybrid Proofs Split zkRollup Logic and Hash Heavy Checks

The short version

Hong Kong University of Science and Technology and OKX researchers proposed Gryphes, a hybrid zero knowledge stack for rollups. It mixes matrix lookup arguments with SNARK friendly hash gadgets. The goal is fast proofs that still support many transaction types.

Rollup teams want proofs that stay small on chain. They also want freedom to add new app logic. Gryphes tries to split those jobs instead of forcing one proof system to do everything.

What the researchers did

The team posted Gryphes: Hybrid Proofs for Modular SNARKs with Applications to zkRollups on the IACR Cryptology ePrint Archive as 2026/596 (preprint, not yet peer-reviewed). They argue real rollup workloads split into flexible app logic and heavy hash heavy cryptography. Gryphes links a Plonk style matrix lookup track with commit and prove SNARKs tuned for hashes. A public artifact repository supports the PETS 2026 style benchmarks described in the paper.

The paper walks through how proofs from each track are linked. App steps that fit matrix lookup stay on that path. Signature checks and membership proofs route to hash tuned SNARKs. The authors describe Groth16 use for signatures and RSA style accumulators for membership.

Benchmark scripts in the artifact repo mirror figures in the write up. That helps other cryptographers rerun timing tables on similar hardware.

The ePrint lists security goals for each module. Readers can see which proof backs which claim on chain.

What they found

The authors integrate Groth16 for signatures and RSA style accumulators for membership checks. They report a zkRollup prototype with constant size proofs and support for thousands of transaction types in their benchmark story. Matrix lookup plus Plonk is meant to cut translation overhead compared with forcing all logic through a single general purpose zkVM path.

They present proof size and prover time trends as transaction mix changes. Hash heavy paths dominate cost in many rollup designs. Isolating those paths is the main win the paper claims.

Security claims still rest on the new linking layer. Independent review has not yet stress tested every gadget choice.

Why it matters for you

Layer two teams must balance proof cost, developer flexibility, and audit surface. Hybrid designs that isolate hash heavy paths may reduce prover time without blocking new opcodes. Exchanges and wallet vendors should track whether such modular proofs become production templates or stay research demos.

Protocol engineers can compare Gryphes to monolithic zkVM rollups on their own opcode mix. If most gas is hashes and signatures, a split design may shave prover hours. Wallet teams care because slower proofs can delay withdrawals users see.

Auditors get a clearer map of which SNARK backs which feature. That can shrink review scope when only one module changes.

What to keep in mind

This is a preprint, not yet peer-reviewed. New linking proofs need independent cryptanalysis. Production rollups still depend on bridge and sequencer trust assumptions that proofs alone do not remove.

Artifact code is not a mainnet deployment. Parameter choices in demos may not match your chain’s curve or hash suite.

Run your own prover benchmarks on the mix of txs your chain actually sees. Paper times on lab hardware may not match your cloud prover fleet.

Wait for outside review of the link proofs before you fork mainnet code. A hybrid stack adds moving parts that auditors must trace end to end.

Dev teams should track PETS 2026 talks tied to the artifact repo for any post print fixes.

The ePrint id is 2026/596. File that id in your zk research tracker next to other rollup proof notes from 2026.

Compare proof sizes in the paper to your current on chain verify gas before you promise users faster withdrawals.

Photo: Pexels

Key takeaways

• Gryphes splits app logic and cryptographic cores across linked proof systems.
• The ePrint describes a zkRollup prototype with constant size proofs.
• Teams should wait for peer review before treating security as settled.

Sources

• Primary — IACR ePrint, Gryphes: Hybrid Proofs for Modular SNARKs with Applications to zkRollups (2026). Defines the hybrid proof architecture and rollup prototype claims.
• Secondary — GitHub, hkust-okx-zkdex-project/pets26-artifacts (2026). Artifact code linked from the paper page.